Desmoglein-2-Integrin Beta-8 Interaction Regulates Actin Assembly in Endothelial Cells: Deregulation in Systemic Sclerosis

Background The inability of endothelial cells of patients affected by the diffuse form of Systemic sclerosis (SSc) to perform angiogenesis is a marker of the disease. We previously demonstrated that desmoglein-2 reduction is a major difference between (SSc)-microvascular endothelial cells (MVECs) and normal (N)-MVECs. Here we investigated the role of desmoglein-2 in human N-MVECs and SSc-MVECs angiogenesis. Methodology/principal findings Angiogenesis was studied by Matrigel invasion, capillary morphogenesis in vitro and Matrigel plug assay in vivo. Gene profiling was studied by Affymetrix technology and signal transduction by Western blotting. Colocalization was validated by immunoprecipitation and confocal microscopy. SiRNAs were used to validate the roles of specific molecules. We observed that desmoglein-2 co-localizes with integrin-beta8 in N-MVECs. This complex is required to signal through Rac, FAK, SMAD1/5 and MAP-kinases, promoting an angiogenic program. Inhibition of desmoglein-2 by DSG2-siRNA impaired actin stress fibres formation, capillary morphogenesis in vitro and angiogenesis in vivo. Transcriptome profiling after DSG2 inhibition revealed alterations of several genes involved in actin organization. siRNA inhibition of integrin-beta8 and RAC2 also resulted into capillary morphogenesis impairment in N-MVECs, due to reduced expression of the same actin-assembly genes that were down-regulated by DSG2 silencing. SSc-MVECs showed down-regulation of the same genes in DSG2-siRNA treated N-MVECs, suggesting that impairment of desmoglein-2/integrin-beta8 complex contributes to angiogenesis derangement in SSc. Transfection of DSG2 in SSc-MVEC partially restored their angiogenic properties in vitro. Conclusions/significance We have shown that impairment of actin assembly as a result of desmoglein-2/integrin-beta8 complex formation is a major factor contributing to angiogenesis deregulation in Systemic sclerosis.


Introduction
We have previously studied Systemic sclerosis-microvascular endothelial cells (SSc-MVECs) as a model system to identify alterations accounting for anti-angiogenesis [1][2][3][4]. We observed that SSc-MVECs, which over-express pro-angiogenic factors, also over-produce anti-angiogenic molecules and lack operative systems required to perform angiogenesis. Reduced expression of desmoglein-2 (DSG2) in SSc-MVECs was one of the most striking differences, as demonstrated by differential transcriptome profiling and by immunohistochemistry of endothelial cells (EC) performed in patients affected by the diffuse form of SSc [3]. Although desmoglein-1/2 has been identified as a structural component of EC [5,6], and has been described in many transcriptome-profiling studies of EC [7][8][9], its function in EC and in particular in angiogenesis has not yet been studied. Desmoglein-2 belongs to the family of desmosomal cadherins, involved in many biological processes including cell adhesion, morphogenesis, cytoskeletal organization and cell sorting/migration, as well as in pathological conditions such as cancer [10,11]. The most studied cadherin of EC is vascular endothelial (VE)-cadherin (CDH5) which plays an important role in vasculogenesis and vascular remodelling [12], although its absence still allows EC to assemble to form vascular networks, indicating that VE-cadherin is not essential in this process [13]. Small GTPases of the Rho family regulate cadherinbased cell adhesion and linkage of cadherins to the actin cytoskeleton, which is responsible for strong adhesion [10]. The intracellular regions of desmosomal cadherins, such as desmoglein-2, interact with plakoglobin and with plakophilin, which in turn links to desmoplakin and consequently with the intermediate filament network. Plakophilin induces filopodia, reduces cell contacts and stimulates the formation of motility-associated structures. The association of plakophilin at the crossroads of adhesion and motility points to an important role of desmoglein in migration, wound healing and tissue formation through its possibility to regulate actin dynamics [14]. Since ECs contain a large number of filopodia, which sense gradients of guidance cues and consequently turn the cell towards chemoattractants and angiogenic factors, we have studied by siRNA-dependent loss of function the role of desmoglein-2 in actin filament assembly in N-MVECs.
To identify the biological processes and pathways involved in the anti-angiogenic phenotype due to the reduced expression of desmoglein-2, we performed gene expression profiling of N-MVECs before and after siRNA silencing of the DSG2 gene and investigated the alterations at the mRNA, protein and functional levels. Further, we have induced a desmoglein-2 gain-of-function in SSc-MVECs, observing a substantial increase in their angiogenic capabilities.

Ethics Statement
For animal studies the local Institutional Animal Care and Use Committee of the Medicine Faculty of Florence (Ospedale di Careggi) and the Italian Ministry of Health (Ministerial Decree n 21/2010, released on January 28, 2010) approved the experimental protocols described in the study. All surgery was performed under sodium pentobarbital anesthesia, and all efforts were made to minimize suffering. For human skin biopsies the local Ethical Committee of the Medicine Faculty of Florence approved the study protocols and participants provided their written informed consent to participate in this study.

Subjects, tissue biopsies, endothelial cells
MVECs were isolated from 3 normal subjects (N-MVEC) and 3 SSc patients (SSc-MVEC) affected by the diffuse form of the disease. For patient selection and EC isolation refer to references 1-3. Where present, cell colonies were detached with EDTA, CD31-positive cells were subjected to immuno-magnetic isolation with Dynabeads-CD31 (Dynal Biotech) and characterized as described [1][2][3]15]. Cells were maintained in complete EC-growth medium (ECGM) [1] and used between the 3 rd and 7 th passage in culture. Both N-MVECs and SSc-MVECS were screened for endothelial markers (CD31, KDR, CD105, vWF and ULEX lectin) at regular intervals, showing that the EC markers profile did not significantly change from the 3 rd to the 10 th passage.

si-RNA treatment of N-MVECs
Targeting and not-targeting siRNAs were obtained from Dharmacon. Specific silencing of selected genes (DSG2, ITGB8, RAC2) was performed by transfection of N-MVECs with smallinterfering-RNA (siRNA) (SMART-siRNA-pool, each pool targeting a single gene: human DSG2, ITGB8, RAC2, respectively), according to the manufactures's instruction. Not-targeting si-RNA pool constructs were used as negative control (siCONTROL). To validate gene silencing, the relevant mRNA levels were determined by a quantitative real-time (RT)-PCR, as described in the dedicated paragraph. To favour internalization into N-MVECs, siRNAs were incorporated into cationic liposomes, utilizing DharmaFECT transfection reagent. Cells were incubated with transfection mix (24-48 h for mRNA analysis and 48 h for protein and phenotypic analysis, respectively).

RNA extraction and Microarray Analysis
Total RNA was isolated from N-MVECs and SSc-MVECs using RNeasy Kit (QIAGEN). Gene expression profiles of N-MVECs and siDSG2-N-MVECs were obtained by Affymetrix technology and Human Genome U133 Plus 2.0 GeneChip, that contain 54,675 probe-sets allowing evaluation of the expression of 47,000 transcripts and variants. In order to prepare RNA target for gene expression analysis, the GeneChipH 3'IVT Express Kit was used (http://www.affymetrix.com). After scanning, data files were checked for quality parameters. Microarray analysis was performed according to Affymetrix suggestions (GeneChip Expression Analysis: data analysis fundamentals in http://www. affymetrix.com) and our previous experience [3,[16][17][18].

Statistical analysis and functional classification of microarray data
Data have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo) and are accessible through GEO Series accession number GSE21547 (http://www. ncbi.nlm.nih.gov/geo/query/acc.cgi? acc = GSE21547). Image and expression data files were generated with Affymetrix MAS 5.0. Low level and statistical analysis was done using R 2.9. Microarray data were first processed in R environment (http:// www.r-project.org) by Affy-package to identify present/absent probe-set, and then subjected to a normalization step according to the Micro Array Suite (MAS) method (www.affymetrix.com, www. bioconductor.org) [19]. In order to identify differentially expressed genes in siDSG2-N-MVECs with respect to N-MVECs, we applied a t-statistic variant approach. We used the significance analysis of microarrays (SAM) method [20] by Siggenes package (www. bioconductor.org) [21].
Functional classification of the differentially expressed genes was done using Database for Annotation, Visualization and Integrated Discovery (DAVID) (http://david.abcc.ncifcrf.gov/) [22]. DAVID analysis was performed including the Gene Ontology databases, three pathway databases (Biological Biochemical Image Database, KEGG PATHWAY and BIOCARTA) and three functional categories (SP_PIR_KEYWORDS, COG_ONTOLOGY and UP_SEQ_FEATURE). The differentially expressed genes were used for cluster analysis by a Ward hierarchical clustering algorithm separately to samples and genes. To cluster samples we used the matrix of the Pearson's correlation coefficient, while for genes we used the matrix of the Euclidean distance. The cluster analysis and the heat-map were both performed using the R statistical environment.

Real-Time PCR
To validate silencing experiments and microarray data, we tested 12 genes by quantitative RT-PCR using TaqMan ABI-PRISM 7900 (Applied Biosystems): DSG2 gene, 4 genes with increased expression and 7 genes with decreased expression at microarray experiments. We used 25 ng of cDNA for each sample. The ''Deltadelta method'' was used for comparing relative gene expression results (Applied Biosystems). Expression of target genes was normalized to GAPDH and displayed as fold-change or log 2 fold  The table summarizes selected genes that had a log2 expression greater than X and P values less than Y, chosen for their biological relevance in cytoskeleton assembly and angiogenesis. Expression: up arrows = genes with increased expression in siDSG2-N-MVECs; down arrows = genes with decreased expression in siDSG2-N-MVECs. * = genes resulted differentially expressed by more than 1 gene set. Genes chosen for real-time PCR validation or siRNA-dependent silencing are reported in bold characters. doi:10.1371/journal.pone.0068117.t001 change relative to control RNA used as the calibrator. In Table S1 the tested genes and their assay ID are reported.

Invasion assay and proliferation
MVECs invasion through Matrigel-coated porous filters was evaluated by the Boyden chamber [1]. 8610 3 MVECs were placed in the upper chamber in culture medium containing 2%

Cell Viability Assay
The viability N-MVECs under each relevant condition was determined by a cell proliferation assay using WST-1 reagent (Roche). WST-1 is a water-soluble sulfonated tetrazolium salt that is cleaved by cellular succinate-dehydrogenases in living cells, yielding dark blue formazan. Damaged or dead cells exhibit reduced or no dehydrogenase activity. Briefly, N-MVECs were plated onto a 96-multiwell plate in quadruple and treated with siRNAs as described above. After 72 hours WST-1 solution/ culture medium (5 mmol/l, 1:9) was added to each well. Following 2-hours incubation at 37uC, absorbance at 450 nm (reference at 630 nm) was measured by a Multiskan JX microplate reader. Percentage of cell viability was calculated based on the absorbance measured relative to that of the untreated control cells maintained in culture medium alone.

Matrigel-sponge-assay in mice
Aliquots of 50 ml of serum-free ECGM, containing 50 U/ml heparin and 600 ng/ml VEGF, with or without DSG2-siRNA, combined with DharmaFect, were added to un-polymerized Matrigel at 4uC at a final volume of 0.6 ml. Matrigel suspension was injected subcutaneously into flanks of C57/BL6 male mice (Charles River) using a cold syringe. At body temperature Matrigel polymerizes and becomes vascularized within 4 days in response to angiogenic substances. Groups of 4 pellets were injected for each treatment. Pellets were removed, photographed with a stereo-microscope, minced and diluted in water to measure hemoglobin content with a Drabkin reagent kit (Sigma Aldrich).

In vitro capillary morphogenesis assay
Matrigel (0.5 ml; 10-12 mg/ml) was pipetted into 13-mm tissue culture wells and polymerised for 30 min to 1h at 37uC [4]. MVECs were plated (60610 3 /ml), in complete MCDB medium, supplemented with 30% FCS, and 20 mg/ml EC-growth-supplement (ECGM). Morphogenesis was evaluated after 6 h with an  [23]. We used the same operating mode of image analysis: Manders overlap coefficients indicate an overlap of the signals and thus represent the degree of colocalization between the red and green pixels: their values range from 0 (no overlap) to 1 (complete overlap). The background signal on each image was initially corrected using the Image J background subtraction function and, whenever possible, single cells on the images were selected using the Iasso tool (which defines a so-called ''region of interest''). Colocalization was then calculated, after choosing the threshold values for the green and red channels, with the aforementioned plugin on the regions of interest previously defined. Quantization of Desmoglein-2 fluorescence and Integrin-beta8 fluorescence was conducted on z stacks of 70 sections corresponding to a thickness of approximately 0.8 mm passing through the middle of the cells. Regions of interest were manually drawn around each cell, and their integrated intensity after background correction was measured in Desmoglein-2 and Integrin-beta 8 channels. Image J software was used for fluorescence quantization, and Origin 6.1, Version 6.1052 (B232) was used for statistical analysis. Results are expressed as mean 6 SD. Multiple comparisons were performed by 1-way ANOVA with Bonferroni correction. A P value ,0.05 was considered statistically significant. Some immunofluorescence experiments with TRITC-labeled phalloidin, aimed to show the involvement of specific signalling pathways in actin polymerization, were performed in the presence of the following signalling inhibitors: FAK inhibitor 14 (Santa Cruz Biotech), MEK inhibitor UO126 (Promega), p38 inhibitor SB202190 and TGFb receptor type I and type II inhibitor LY2109761 (both from SelleckBio.com).

Transient transfection of SSc-MVECs with DSG2
For transient transfection of the human DSG2 gene, SSc-MVECs (6610 6 cells) were incubated in 500 ml of serum-free OPTI-MEM (Life Technologies) with the addition of recombinant vector (CMV6-XL DSG2) or pCMV6-XL alone as empty vector control (15 mg each) (OriGene Technologies) and then were electrophorated using a Bio-Rad Gene Pulser apparatus.

Statistical analysis
Results are expressed as mean 6 SD. Comparisons were performed by the Student test, differences were considered statistically significant at p,0.05.

DSG2 silencing in N-MVECs and its effects on in-vitro parameters of angiogenesis and in vivo Matrigel sponge assay
After silencing of DSG2 in N-MECs (siDSG2-N-MVECs), DSG2 expression levels were reduced about 14-fold with respect to control cells ( figure 1A). This correlated with a decrease in

Affymetrix Gene Expression Profiling
To investigate the biological processes and pathways altered by reduced expression of DSG2, Affymetrix gene expression profiling of N-MVECs before and after siRNA silencing of DSG2 was performed. Throughout all gene expression profiling experiments N-MVECs 1, 2 and 3 were used at passages 4, 4 and 5, respectively in order to minimize changes of gene expression related to cell passaging.

Classification of differentially expressed genes and cluster analysis
We performed the functional classification of the 2,945 differentially expressed genes using DAVID analysis and including in the analysis information from the Gene Ontology database (Molecular Function, Biological Process and Cellular Component), three pathway databases (Biological Biochemical Image Database, KEGG PATHWAY and BIOCARTA) and three functional categories (SP_PIR_KEYWORDS, COG_ONTOLOGY and UP_SEQ_FEATURE). Several classes were significantly associat-ed with siDSG2-N-MVECs. Among the over-represented annotation terms, we observed a high number of functional terms and pathways implied in cytoskeleton organization and biogenesis, angiogenesis, blood vessel development. These data, together with those of figure 1, suggested that cytoskeleton deregulation could be a crucial mechanism of the altered phenotype of siDSG2-N-MVECs. Table 1 shows a selection of differentially expressed genes involved in cytoskeleton biogenesis/organization and angiogenesis.
The 2,945 differentially expressed genes were then used for cluster analysis (figure S1, dendrogram). The gene expression profiles clearly distinguished between siDSG2-N-MVECs and N-MVECs and the cluster analysis identified six groups of genes by similar expression levels (identified as box A-F in figure S1). Interestingly, in the cluster E, one of the gene cluster with the higher expression levels, we found many genes significantly associated with classification terms involved in cytoskeleton, cell motility and migration, and angiogenesis (table S2).

Altered actin cytoskeleton assembly and Rho/Rac-GTPase activity in silenced N-MVECs
To verify whether DSG2 silencing altered actin assembly, thus accounting for inhibition of angiogenesis in vitro and in vivo, we visualized actin organization in control and siDSG2-N-MVECs, as related to SSc-MVECs. Actin stress fiber disassembly was found in siDSG2-N-MVECs (figure 2A) and a similarly altered actin organization like that of SSc-MVECs. Since small GTPases of the Rho/Rac family mediate cadherin-dependent cytoskeletal rearrangements and cell motility [10], as well as integrin-mediated outside-in signals to actin cytoskeleton, we studied Rho/Rac activation by Western blotting. N-MVECs and SSc-MVECs both express Rac (figure 2B), whose activated form, evident in N-MVECs, resulted barely detectable in both siDSG2-N-MVECs and in SSc-MVECs. Interestingly, DSG2-silenced N-MVECs expressed lower amounts of Rac, in line with data obtained by microarray in siDSG2-N-MVECs, while no modification of Rho was observed (figure 2B).

Real Time-PCR validation of microarray data
By real time-PCR, we assessed the expression of 12 selected differentially expressed genes: seven genes with a decreased expression (MACF1, which stabilizes actin at sites where microtubules and microfilaments meet; DIAPH1, DIAPH2 and ARPC3, involved in regulation of actin polymerization; CDH5, whose intracellular partners a-catenin, b-catenin and vinculin interact with the actin network; the integrin ITGB8, which mediates cell-cell, cell-ECM interactions and cytoskeleton assembly in ECs; RAC2, involved in endothelial integrin signalling, and DSG2. Four genes with an increased expression (VCL, a cytoskeletal protein involved in anchoring F-actin to the membrane; THBD, a marker of endothelial injury; THBS1, an angiogenesis inhibitor; MID1, a microtubuleassociated protein) were assessed in parallel ( Figure 2C). Down regulation of the actin-cytoskeleton regulating genes by silencing of ITGB8 and of RAC2 in N-MVECs We examined by real time-PCR the expression of MACF1, DIAPH1, DIAPH2, ARPC3, RAC2, CDH5, ITGB8 and DSG2 following siRNA-dependent silencing of either ITGB8 or RAC2 ( Figure 2D). The data show that silencing of either gene provided a gene expression pattern similar to that obtained upon silencing of DSG2 (shown in figure 2C). These results, coupled with those shown in figure 2B, were strongly suggestive of a possible functional and spatial relationship of the protein products of DSG2, ITGB8 and RAC2 genes.

Colocalization of integrin-beta8 and desmoglein-2 in N-MVECs and related transduction pathways
Because in contrast to integrins and classical cadherins, desmogleins are not known to elicit intracellular signalling [24], and because our results indicated that DSG2 and ITGB8 silencing produced angiogenesis inhibition and a similar pattern of expression in genes regulating actin in N-MVECs, we evaluated whether integrin-beta8 and desmoglein-2 colocalize in MVECs, in order to verify whether integrin-beta8 was present as a transduction partner in a putative desmoglein-2/integrin-beta8 complex in MVECs. Confocal Laser Scanning Microscopy was used to colocalize integrin-beta8 and desmoglein-2 by specific antibodies. Both antibodies provided a diffuse micro and macro-dotted distribution, together with a slight membrane localization that, upon merging, revealed an almost complete co-localization (figure 3A). Co-localization was validated by immuno-precipitation experiments (figure 3B). Previous data had suggested that the integrin-beta8 cytoplasmic domain did not interact with the cytoskeleton and with cytoplasmic signalling pathways in an adhesion-promoting fashion [25]. However, it still appeared to be involved in late events in shaping cell morphology, such as process extension, indicating integrin-beta8 as a ''divergent'' signalling integrin [25]. Integrin-beta8 can signal through at least two different pathways: a classical FAK-mediated pathway and a TGFb-mediated one. The latter pathway relies on the property of integrin-beta8 to release ECM-entrapped TGFb in an activated form [26]. We therefore examined a series of transduction pathways related to classical integrin-mediated signalling and to TGFb-mediated signalling. Both FAK phosphorylation and TGFb-dependent SMAD1/5 phosphorylation decrease upon silencing of DSG2 or ITGB8 (figure 4A). The main components of down-stream transduction MAPK signalling pathway, ERK1/2 and p38a, which control gene expression, also show decreased phosphorylation upon DSG2 or ITGB8 silencing. The functional role of these signalling pathways in the regulation of actin assembly in MVECs was confirmed by the use of specific inhibitors of FAK, MEK, p38 and TGFbRI/II (figure 4).
Biological consequences of the silencing of the 7 selected genes with reduced expression in siDSG2-N-MVECs In order to verify the role of the genes with reduced expression in siDSG2-N-MVECs, we performed the silencing of 7 selected genes: MACF1, DIAPH1, DIAPH2, ARPC3, RAC2, CDH5, and ITGB8, and evaluated the biological consequences of silencing in terms of capillary morphogenesis on Matrigel and stress fiber organization. ITGB8 silencing produced the inhibition of vessellike structures and disrupted the actin cytoskeleton assembly, as shown by stress fiber disorganization with respect to untreated and control siRNA-treated MVECs (figures 5A and B). Silencing of the other genes gave similar results (data not shown).

Evaluation in actin cytoskeleton regulating genes in SSc-MVECs and forced expression of DSG2 in SSc-MVECs
Our previous studies on transcriptome profiling of SSc-MVECs [3] tested 14,000 transcripts, as compared to the 47,000 transcripts of this study. The previous profiles included DSG2 but did not assay the other genes involved in actin-cytoskeleton organization investigated and validated here. To investigate whether siDSG2-N-MVECs and SSc-MVECs shared alterations of the actincytoskeleton genes down-regulated by siDSG2, we assessed by RT-PCR the expression of MACF1, DIAPH1, DIAPH2, ARPC3, RAC2, CDH5 and ITGB8 genes in three SSc-MVEC lines ( figure 6A). Our results showed that these genes had reduced expression in all the SSc-MVEC lines as compared with N-MVECs. To test the hypothesis that desmoglein-2 is required for efficient angiogenesis of N-MVECS that is lacing in SSc, we transiently transfected the three SSc-MVEC lines with DSG2. Immunofluorescence analysis of DSG2-transfected SSc-MVECs showed significantly increased levels of desmoglein-2 and integrinbeta 8 expression along with significantly increased desmoglein-2/ integrin-beta 8 colocalization ( figure 6B and table 2). Further, the DSG2 transfected SSc-MVECs partially recovered the capability to undergo capillary morphogenesis in vitro ( figure 6C).

Discussion
We have demonstrated here that desmoglein-2 and integrin-beta8 are colocalized in N-MVECs and that the complex is required to deliver signals, mediated by the small GTPase Rac, by FAK and SMAD 1/5 phosphorylation, and by phosphorylation of the MAP-Kinase ERK 1/2 and p38a, which promote angiogenesis in N-MVECs. Inhibition of DSG2 and/or ITGB8 expression by siRNA impaired N-MVECs angiogenesis, inducing actin stress fibres disassembly. The effect depended on down-regulation of genes involved in actin assembly, such as ARPC3, CDH5, DIAPH1, DIAPH2 and MACF1. Inhibition of each one of such genes produced actin disassembly and capillary morphogenesis impairment. The same molecules were also down-regulated in SSc-MVECs, suggesting that impairment of the desmoglein-2/ integrin-beta8 complex contributes to angiogenesis derangement in the diffuse form of Systemic sclerosis.
Although desmoglein-1/2 has been identified as a structural component of EC intercellular junctions [5], and has been reported in transcriptome-profiling studies of ECs [3,[7][8][9], its role in the context of EC function has never been experimentally addressed. Our data identify unexpected properties of DSG2 in regulation of actin dynamics in ECs.
Reducing DSG2 expression with siRNA in N-MVECs to levels comparable with SSc-MVECs [3] produced a similar inhibition of angiogenesis in vivo and in vitro. The transcriptome-profiling of DSG2-silenced N-MVECs identified a number of differentially expressed genes associated with cytoskeleton organization, cell motility and migration, and angiogenesis, suggesting that cytoskeleton deregulation could be a crucial mechanism of the altered angiogenic phenotype of siDSG2-N-MVECs. In particular, in siDSG2-N-MVECs ARPC3, which promotes actin polymerization [27], had decreased expression suggesting a possible role in cytoskeleton disorganization. The protein encoded by the MACF1 gene belongs to the plakin family of cytoskeletal linker proteins [28] and appears to stabilize actin at sites where microtubules and microfilaments meet. Actin polymerization also requires proteins known to interact with diaphanous protein DIAPH1 and DIAPH2, that work through RAC activation [29]. Functioning as a classic cadherin, the protein encoded by CDH5 gene plays an important role in EC biology through organization of the intercellular junctions and induction of actin assembly [10][11][12][13]. The ITGB8 gene is a member of the integrin-beta family and encodes a singlepass type I membrane protein [30]. In general, integrin complexes mediate cell-cell and cell-extracellular-matrix interactions, as well as cytoskeleton organization [25]. The GTPase RAC2 is involved in endothelial integrin signaling and in the postnatal neovascularization response in vivo [31]. Moreover, it has been shown that the small GTPase RAC2 regulates actin structures in a variety of cells [32]. Down-regulation of this same set of genes was obtained by siRNA-dependent inhibition of DSG2, ITGB8 or RAC2, suggesting that an outside-in informational flux dependent on their expression could be inhibited regardless of the silenced gene.
Because desmogleins are not known to elicit intracellular signalling [24], and ITGB8 silencing resulted into angiogenesis inhibition and down-regulation of actin-regulating genes in a pattern similar to that elicited by DSG2 silencing in N-MVECs, we hypothesized that integrin-beta8 was present as a transduction partner in a putative integrin-beta8/desmoglein-2 complex. This hypothesis was confirmed by co-localization of integrin-beta8 and desmoglein-2 in N-MVECs by Confocal Microscopy and immunoprecipitation. In particular, by confocal microscopy we have identified a non-conventional distribution of desmoglein-2, characterized by a diffuse micro and macro-dotted pattern over the cell surface in N-MVECs. This is not surprising, considering that desmosomes are not present in endothelial cells and that desmoglein-2 has been described as a novel solitary surface glycoprotein that is present in non-junction-restricted forms in cutaneous cells devoid of desmosomes, upon malignant transformation [33].
We also observed a complex array of signaling pathways that are likely to cooperate in integrin-beta8/desmoglein-2-dependent angiogenesis. Since integrin-beta8 is the signalling molecule of the integrin-beta8/desmoglein-2 complex, inhibition of signalling dependent on DSG2 silencing is likely to be ascribable to a possible function of desmoglein-2 as an ''adjuvant'' molecule required for integrin-beta8 to exert its signalling function. The pathways involved range from FAK to the small GTPase RAC, the MAP-Kinase system, considered immediately upstream of gene expression, and, surprisingly, the TGFb-dependent ECrestricted SMAD1/5 transduction pathway. An intimate relationship between some integrins and the activation of latent TGFb1 has recently been described [26]. Integrin-dependent activation of latent TGFb may disclose the full range of TGFb activities, including angiogenesis promotion [34][35][36]. It is interesting to note that upon DSG2 silencing in N-MVECs, down-regulation of Rac and of its activation is not counterbalanced by Rho activation. From a general point of view, motile cells assume several phenotypes, including a mesenchymal mode, in which the cells are elongated and fibroblast-like, and a distinct amoeboid mode, with less adherent properties and extensive membrane deformation. The so-called ''mesenchymal-type movement'', which heavily depends on cell surface-associated proteases, has been shown to depend on activation of Rac and inhibition of Rho GTPases. In contrast the cell movement referred to as ''amoeboid-like'' has been associated with the loss of cell proteases activity and is characterized by an opposite phenotype of small GTPases [37,38]. Therefore, a high Rho/Rac ratio identifies an amoeboid motility, while a low Rho/Rac ratio connotes a mesenchymal motility. However, the Rho/Rac balance always requires a proper functioning of the actin cytoskeleton, which is the most important pre-requisite in order to shift from a mesenchymal to an amoeboid movement style. Since the actin assembly genes are deregulated in the absence of desmoglein2/integrin-beta8 complex, si-DSG2treated N-MVECs are unable to shift to an amoeboid movement, which may account for the absence of Rho activation.
Our results contribute to add new insights into the mechanisms of angiogenesis deregulation in Systemic sclerosis. In addition to DSG2, also MACF1, DIAPH1, DIAPH2, ARPC3, RAC2, CDH5, and ITGB8 genes showed a reduced expression in SSc-MVECs, suggesting that the very poor expression of DSG2 previously observed in our study of the differential gene expression profiling of N-MVECs and SSc-MVECs [3] may play a critical role in the pathogenesis of Systemic sclerosis angiogenesis impairment.

Conclusions
Our data support a key role of the integrin-beta8/desmoglein-2 complex in the regulation of the angiogenic process. In agreement with the observations of this study, transient transfection of SSc-MVECs with DSG2 restored the loss of co-localization between desmoglein-2 and integrin-beta8, allowing at the same time a partial rescue of their capillary morphogenesis ability in vitro. These data show that DSG2, a recently emerged component of ECs, is particularly important in the regulation of a correct angiogenesis process and that its loss plays a relevant role in the complex scenario of angiogenesis deregulation in Systemic sclerosis. Figure S1 Hierarchical cluster analysis of the 2,945 differentially expressed genes in N-MVEC subjected to DSG2 silencing with respect to N-MVEC.

(TIF)
Table S1 Gene Name, Gene Symbol and Assay ID (Applied Biosystems) of genes analyzed by real time PCR.

(DOC)
Table S2 Over represented terms after functional classification (DAVID analysis) of the genes found in the cluster E of Supplemental Figure S1, associated with cytoskeleton biogenesis and organization and angiogenesis.