The Differentiation-dependent Desmosomal Cadherin Desmoglein 1 Is a Novel Caspase-3 Target That Regulates Apoptosis in Keratinocytes*

Although a number of cell adhesion proteins have been identified as caspase substrates, the potential role of differentiation-specific desmosomal cadherins during apoptosis has not been examined. Here, we demonstrate that UV-induced caspase cleavage of the human desmoglein 1 cytoplasmic tail results in distinct 17- and 140- kDa products, whereas metalloproteinase-dependent shedding of the extracellular adhesion domain generates a 75-kDa product. In vitro studies identify caspase-3 as the preferred enzyme that cleaves desmoglein 1 within its unique repeating unit domain at aspartic acid 888, part of a consensus sequence not conserved among the other desmosomal cadherins. Apoptotic processing leads to decreased cell surface expression of desmoglein 1 and re-localization of its C terminus diffusely throughout the cytoplasm over a time course comparable with the processing of other desmosomal proteins and cytoplasmic keratins. Importantly, whereas classic cadherins have been reported to promote cell survival, short hairpin RNA-mediated suppression of desmoglein 1 in differentiated keratinocytes protected cells from UV-induced apoptosis. Collectively, our results identify desmoglein 1 as a novel caspase and metalloproteinase substrate whose cleavage likely contributes to the dismantling of desmosomes during keratinocyte apoptosis and also reveal desmoglein 1 as a previously unrecognized regulator of apoptosis in keratinocytes.

Although a number of cell adhesion proteins have been identified as caspase substrates, the potential role of differentiation-specific desmosomal cadherins during apoptosis has not been examined. Here, we demonstrate that UV-induced caspase cleavage of the human desmoglein 1 cytoplasmic tail results in distinct 17-and 140-kDa products, whereas metalloproteinase-dependent shedding of the extracellular adhesion domain generates a 75-kDa product. In vitro studies identify caspase-3 as the preferred enzyme that cleaves desmoglein 1 within its unique repeating unit domain at aspartic acid 888, part of a consensus sequence not conserved among the other desmosomal cadherins. Apoptotic processing leads to decreased cell surface expression of desmoglein 1 and re-localization of its C terminus diffusely throughout the cytoplasm over a time course comparable with the processing of other desmosomal proteins and cytoplasmic keratins. Importantly, whereas classic cadherins have been reported to promote cell survival, short hairpin RNA-mediated suppression of desmoglein 1 in differentiated keratinocytes protected cells from UV-induced apoptosis. Collectively, our results identify desmoglein 1 as a novel caspase and metalloproteinase substrate whose cleavage likely contributes to the dismantling of desmosomes during keratinocyte apoptosis and also reveal desmoglein 1 as a previously unrecognized regulator of apoptosis in keratinocytes.
Desmosomes are vertebrate cell junctions that anchor the intermediate filament cytoskeleton to the plasma membrane at sites of cell-cell contact and in so doing form a supracellular scaffolding that is essential for maintaining tissue integrity (for review, see Ref. 1). The molecular components of the desmosome fall into three main families: desmosomal cadherins, armadillo family proteins, and plakins. The two types of desmosomal cadherins, desmogleins (Dsgs) 4 and desmocollins (Dscs), are thought to mediate calcium dependent cell-cell adhesion at the membrane, which is strengthened through indirect interactions with the intermediate filament cytoskeleton (for review, see Ref. 2). The armadillo family protein, plakoglobin (Pg), interacts directly with the cytoplasmic tail of the desmosomal cadherins (3)(4)(5)(6), thereby connecting the transmembrane glycoproteins to the obligate desmosomal protein, desmoplakin (DP) (7). DP links the cell surface to the cytoskeleton by associating with Pg at its N terminus and intermediate filaments through its C terminus (8 -10). Plakophilins, additional members of the armadillo family, can interact with desmosomal cadherin cytoplasmic domains (11,12), enhance recruitment of DP to the membrane, and likely facilitate lateral clustering of the plaque components to enhance the mechanical strength of the junction (13,14).
Four isoforms of Dsgs and three isoforms of Dscs have been identified in humans (15)(16)(17). The cytoplasmic domains of all of the Dsg isoforms contain regions of unknown function, including a unique repeating unit domain (RUD). Each Dsg isoform includes a RUD containing different numbers of a 29-amino acid repeat (16 -18). Dsg1 is expressed most prominently in the differentiated layers of the epidermis (for review, see Ref. 19). Human autoimmune, genetic, and infectious diseases in which the function of Dsg1 is compromised all support a role for Dsg1 in maintaining epithelial tissue integrity (for review, see Ref. 20).
Apoptosis is an active, orderly process of programmed cell death that eliminates superfluous, damaged or destructive cells and is critical to the normal development and maintenance of tissue homeostasis in multicellular organisms (21). The process of apoptosis is mediated by caspases, a special family of cysteine proteases that is largely responsible for producing a series of well defined morphological and biochemical changes in apoptotic cells including loss of cell contact and rounding of the cell body, loss of cytoskeletal integrity, and nuclear condensation (22). Caspases involved in the cell death pathway are characterized either as upstream initiators of the apoptotic proteolytic cascade (such as caspase-8) or as downstream effectors (such as caspase-3) and mediate cell death by selectively cleaving critical target molecules at aspartic acid residues within specific recognition sites (for review, see Ref. 23). Of the more than 250 caspase substrates that have been reported thus far (24), many are cell adhesion molecules and cytoskeleton-associated proteins, including gelsolin (25), keratins 18 and 19 (26,27), vimentin (28), p21 activated kinase 2 (29), plectin (30), periplakin (31,32), and platelet endothelial cell adhesion molecule-1 (33). The classical cadherin E-cadherin (34,35) and the desmosomal cadherin Dsg3 (36) together with their associated proteins DP (32,36), plakophilin 1 (36), ␤-catenin (37)(38)(39), and Pg (38 -40) have all been identified as caspase substrates. Targeting of these junction-associated molecules for cleav-age during apoptosis is thought to mediate the orderly dissolution of cell structure and morphology that typifies apoptotic cell death.
As yet, the differentiation-specific desmosomal cadherin, Dsg1, and its partner, Dsc1, have not been investigated as potential caspase targets. In the work presented here, using an in vitro screen for desmosomal protein caspase targets, we show that Dsg1 is cleaved by caspase-3 at a unique site in the RUD domain that is not conserved in other Dsg isoforms. In keratinocytes stimulated to undergo apoptosis, caspaseand metalloproteinase-dependent Dsg1 processing is temporally coordinated with the proteolysis of other desmosomal proteins, resulting in a loss of Dsg1 cell surface localization concomitant with the disruption of cell-cell junctions. Finally, we show that shRNA-mediated suppression of Dsg1 expression in differentiating keratinocytes protects them from UV-induced apoptosis, suggesting a potentially novel function for Dsg1 in facilitating apoptotic elimination of damaged cells.

EXPERIMENTAL PROCEDURES
Generation of cDNA Constructs-A construct encoding the cytoplasmic tail of human Dsg1 with a C-terminal Myc tag (p288) was utilized for in vitro cleavage assays. This construct includes a 5Ј-BamHI/HincII fragment with the signal sequence (nucleotides 24 -276; see Gen-Bank TM accession number AF097935) attached to the Dsg1 membranespanning region and cytoplasmic domains (nucleotides 1740 -3360). A construct encoding full-length, tagged Dsg1 in pBluescript KS (Stratagene) (p281) (3) was used as a PCR template to add a Myc tag, stop sequence, and HindIII site onto the end of the Dsg1 sequence. The p288 construct was generated by digesting p281 with HincII/HindIII and replacing the resulting fragment with the PCR product at the ScaI/ HindIII sites.
A construct encoding a mutant Dsg1 cytoplasmic tail that is missing the obligate aspartic acid residue in the P1 position of the caspase-3 consensus site (p1133) was generated from p288 using the QuikChange site-directed mutagenesis kit (Stratagene). The following oligonucleotide primers were used: 5Ј-GCCTGACTTGCGAGAGGGGTCGAAT-GTTAT-3Ј and 5Ј-CTATAACATTCGACCCCTCTCGCAAGTCA-GG-3Ј. Generation of the mutation was confirmed by automated DNA sequencing.
A full-length Dsg1 construct was altered at the C terminus by PCR addition of a FLAG tag, stop sequence, and BamHI, SnaBI, HindIII sites. A construct encoding full-length Dsg1 in pBluescript KS (p536) was digested with BglII/HindIII, and the resulting fragment was replaced with the PCR product to generate a construct encoding full-length Dsg1, FLAG-tagged at the C terminus in a pBluescript KS vector (p783). The entire FLAG-tagged Dsg1 region was removed from p783 with a BamHI digestion. This was cloned into the BamHI site in pTRE (Clontech) to generate an inducible construct encoding Dsg1.FLAG under the control of the tetracycline-responsive element (p812). Directionality of the insert was confirmed by restriction enzyme digestion.
Antibodies and Reagents-Immunodetection of Dsg1.FLAG was facilitated with antibodies against the C-terminal FLAG tag using either the rabbit polyclonal antibody OctA Probe (Santa Cruz Biotechnology) or a goat anti-FLAG polyclonal antibody (44) (a gift from A. Kowalczyk, Emory University Medical School, Atlanta, GA). Detection of the N-terminal extracellular domain of Dsg1 was carried out using 982, human serum containing autoantibodies against Dsg1 (gift from J. Stanley, University of Pennsylvania Medical School, Philadelphia, PA) or the monoclonal antibody clone P124 (Research Diagnostics, Inc.). The monoclonal antibody 18D4 (gift from M. Wheelock, University of Nebraska Medical School, Omaha, NE) was used to detect the intracellular domain of Dsg1. The C terminus of Dsg1 was detected using the monoclonal antibody 4B2 (45). Pg was detected with an anti-␥-catenin antibody (Transduction Laboratories). DP was detected using the rabbit polyclonal antibodies NW161 (46) and NW6 (47). The rabbit polyclonal antibody 1589 (a gift from R. Oshima, The Burnham Institute, La Jolla, CA) was used to detect keratin 18. ␤-Tubulin expression was detected with the monoclonal anti-tubulin antibody clone Tub 2. 1 (Sigma). A horseradish peroxidase-conjugated mouse monoclonal antibody against the human pro-form of caspase-3 was purchased from BD Biosciences and used to detect pro-caspase-3 and its activation (disappearance). A rabbit polyclonal antibody used to detect glyceraldehyde-3phosphate dehydrogenase was purchased from Novus Biologicals. The mouse monoclonal antibody JL-8 (Clontech) was used to detect GFP. Full-length and cleaved PARP were detected using a mouse monoclonal antibody purchased from BD Pharmingen. Horseradish peroxidaseconjugated anti-mouse, rabbit, goat, and human secondary antibodies were purchased from Kirkegaard and Perry Laboratories, Inc. Alexa Fluor 488 goat anti-human, goat anti-rabbit, and 568 goat anti-mouse IgG antibodies were purchased from Molecular Probes. Doxycycline (Dox) and 4Ј,6-diamidino-2-phenylindole dihydrochloride (DAPI) were obtained from Sigma. The general caspase inhibitor, carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethyl ketone (ZVAD-fmk) was purchased from Enzyme Systems Products. N-(R)-[2-(Hydroxyaminocarbonyl)methyl]-4-methylpentanoyl-L-naphthylalanyl-L-alanine amide (TAPI-0) was purchased from Peptides International. Matrix metalloproteinase (MMP) inhibitor-1 was obtained from Calbiochem. Complete-EDTA-free protease inhibitor pellets were purchased from Roche Applied Science.
Cell Culture-Parental A431 human vulvar epidermoid carcinoma cells were cultured as described (46). The pTet-On plasmid (Clontech) was transfected into these cells using the calcium phosphate precipitation method. Transfectants were selected in 400 g/ml G418 (Mediatech, Inc.). To verify that the drug-resistant lines expressed the tetracycline-responsive transactivator, a luciferase assay (pTRE-Luc, Clontech) was performed (pRL-CMV, Dual-Luciferase Reporter Assay System, Promega). Tetracycline-inducible Dsg1.FLAG lines were established by co-transfecting a G418-resistant single stable line, 68Q71, with a construct encoding Dsg1.FLAG under the control of the tetracycline-responsive element (p812) and a plasmid encoding the puromycin resistance gene (pSV2pac⌬P). Cells were selected in 400 g/ml G418 and 1 g/ml puromycin (Calbiochem). Resistant clones were screened for expression of Dsg1.FLAG by overnight treatment with 4 g/ml Dox. The cells were then lysed in Laemmli sample buffer and subjected to SDS-PAGE and immunoblot analysis. Of the several stable clones generated, one with the most homogenous expression of Dsg1.FLAG, A431/B4, was further characterized in detail and used in the majority of experiments.
An immortalized but not tumorigenic human keratinocyte cell line, HaCaT (48), was cultured as described (49). To induce expression of endogenous Dsg1, cells were grown to confluence in low (0. 07 mM) calcium and then switched to Medium 154 supplemented with human keratinocyte growth supplement, 1000ϫ penicillin/streptomycin/amphotericin B solution (Cascade Biologics), and 1. 2 mM calcium for 6 days.
Phoenix retrovirus producer cells (kindly provided by G. Nolan, Stanford University Medical School, Palo Alto, CA) were cultured and used to generate enhanced GFP and shDsg1 retroviral supernatants as described (45). These supernatants were used to infect HaCaT cells cultured in low calcium Dulbecco's modified Eagle's medium as described for normal human epidermal keratinocytes (45).
Proteolytic Cleavage of the Cytoplasmic Tail of Human Dsg1 in Vitro- 35 S-Labeled wild type or mutant human Dsg1 cytoplasmic tail was prepared by in vitro transcription and translation of DNA constructs encoding nucleotides 24 -276 plus 1740 -3360 of Dsg1 with a C-terminal Myc tag (p288) or the altered form with an aspartic acid to glutamic acid substitution at amino acid 888 (p1133) using the TNT T7 Quick Coupled Transcription/Translation System (Promega). Radiolabeled Dsg1 tail was then incubated with 2.5 or 25 ng of recombinant caspase (1, 2, 3, 5, 6, 7, 8, or 9) for 1 h at 37°C as described (50,51), except 2.5 mM instead of 20 mM dithiothreitol was used here.
Apoptosis Induction and Preparation of Cell Lysates-A431/B4 and HaCaT cells were cultured in 100 or 60-mm tissue culture dishes, and Dsg1 expression was induced. For inhibitor studies, the cells were preincubated with 50 M ZVAD-fmk, 100 M MMP inhibitor-1, or 50 M TAPI-0 for 1 h before induction of apoptosis. The culture media was removed and saved, and the cells were washed once with phosphatebuffered saline (PBS). Dishes were placed with open lids into a UV Stratalinker 1800 (Stratagene) with about 2. 5 inches of 1 bulb (for A431 cells) or 1 entire bulb (for HaCaT cells) available for emission. The A431/B4 cells were subjected to 40 mJ/cm 2 of ultraviolet (UV) radiation, and the HaCaT cells were subjected to 250 mJ/cm 2 to achieve 30 -50% cell death after 24 h. Medium was returned to the dishes, and the cells were harvested at various time points. Floating cells were collected from the media by centrifugation (1000 ϫ g for 5 min). Adherent cells were harvested in Laemmli sample buffer or 1% Triton X-100 lysis buffer (1% Triton X-100, 145 mM NaCl, 10 mM Tris-HCl, pH 7. 4, 5 mM EDTA, 2 mM EGTA, and Complete-EDTA-free protease inhibitor pellet) and combined with collected floating cells for immunoprecipitation or Western blot analysis.
Western Blot Analysis-The total protein concentration of the cell lysates was determined using the Amido Black Assay (52). Equal amounts of total protein in the cell lysates were subjected to SDS-PAGE and transferred onto nitrocellulose membranes. Immunoblotting was performed as previously described (53)  Immunoprecipitation-Floating and adherent cells were collected at various times after induction of apoptosis and pooled. Cells were lysed in 1% Triton X-100 lysis buffer. Cell lysates were cleared by centrifugation, and immunoprecipitations were carried out as described (53).  35 S-labeled wild type human Dsg1 cytoplasmic tail was incubated with control buffer or increasing concentrations of multiple recombinant caspases for 1 h at 37°C. The products of the reaction were analyzed by SDS-PAGE and autoradiography. The full-length human Dsg1 tail (arrows) is cleaved by caspase-7 and preferentially by caspase-3 to produce a 78-kDa product (circles) and a 17-kDa product (asterisks) in vitro. B, 35 S-labeled wild type or D888E mutant Dsg1 cytoplasmic tail was incubated with control buffer or recombinant caspase-3 for 1 h at 37°C. The products of the reaction were analyzed and labeled as above. The D888E Dsg1 cytoplasmic tail mutant is resistant to caspase-3 cleavage at this site. C, schematic representation of the Dsg1 protein with recognized subdomains of the protein noted. The portions of the protein corresponding to the Dsg1 cytoplasmic tail used in the in vitro cleavage assays and the cleavage products generated by its caspase cleavage (78-and 17-kDa products) are indicated. The caspase-3 consensus site ( 885 DLRD 888 ) DLRD is located at amino acids 885-888 of the full-length protein within the RUD. These data are representative of results from multiple independent experiments. C, control buffer; C1-C9, caspase-1-caspase-9; MW, molecular weight; EI-EIV, extracellular domains I-IV; EA, extracellular anchor; M, transmembrane domain; IA, intracellular anchor; ICS, intracellular cadherin segment; L, linker; DTD, desmoglein terminal domain.
Antibodies used include 982 (2 l) and FLAG (40 l M2-agarose affinity gel (Sigma)). Immunoprecipitation from the culture media was performed as above using 2 ml of media and 5 l of 982. The immunoprecipitated proteins were then subjected to SDS-PAGE and Western blot analysis as described above.
Immunofluorescence Microscopy-A431/B4 and HaCaT cells were grown on 22-mm 2 glass coverslips, induced to express Dsg1, then treated or not with UV irradiation. A431/B4 cells were fixed in ice-cold methanol for 2 min. HaCaT cells were fixed in 4% paraformaldehyde for 10 min at room temperature, washed 3 times for 5 min each in PBS, then permeabilized in ice-cold methanol for 2 min. Immunofluorescence/ epifluorescence analysis was performed as described (54) using primary antibodies at the following dilutions: 982 at 1:100; OctA at 1:100; 4B2 at 1:100; NW6 at 1:50. Alexa Fluor anti-human, mouse, or rabbit secondary antibodies were used at a dilution of 1:300. After the final wash the coverslips were incubated for 5 min at room temperature in 2 g/ml DAPI. The coverslips were washed multiple times in PBS and mounted on slides with gelvatol. Images were obtained on a Leitz DMR microscope using a Hamamatsu Orca 100 digital camera and Openlab (Improvision) or Metamorph (Universal Imaging Corporation) imaging software.
Quantitation of Apoptosis-Apoptosis was assessed by nuclear morphology analysis. Dsg1-expressing A431 or HaCaT floating and adherent cells were collected from tissue culture dishes and pooled at various times after exposure to UV radiation. They were fixed briefly in a 1.25% glutaraldehyde solution (Sigma). After several washes in PBS, the cells were stained with 2 g/ml nuclear dye, DAPI, for 30 min at 37°C. The cells were washed and resuspended in PBS. At least 200 nuclei per sample were examined by immunofluorescence microscopy and scored for their fragmented/condensed appearance. The percent cell death in each sample was calculated based on the following formula: no. of apoptotic nuclei/(no. of normal nuclei ϩ no. of apoptotic nuclei) ϫ 100. Each experiment was performed in triplicate.
To assess the effect of Dsg1 knockdown on cell death in HaCaT cells, UV-treated samples were processed for immunofluorescence analysis as above. Assessment of apoptosis was performed as described. The nuclear morphology of each cell was coordinately assessed together with GFP expression and the presence or absence of Dsg1 staining. For control samples, nuclei were assessed in cells that expressed robust levels of Dsg1 and GFP, and for shDsg1/GFP samples nuclei were assessed in cells with GFP expression but little to no Dsg1 expression. Each experiment was performed in triplicate.

Dsg1 Is Cleaved by Caspase-3 at Aspartic Acid 888 in Vitro-We
initially identified the human Dsg1 cytoplasmic domain as a caspase-3 target using an in vitro screen for novel desmosomal substrates of this apoptotic effector enzyme (data not shown). To determine whether this desmosomal cadherin is uniquely targeted by caspase-3, 35 S-labeled Dsg1 cytoplasmic tail was used as a substrate for recombinant caspases-1, 2, 3, 6, 7, 8, and 9. The 95-kDa full-length Dsg1 cytoplasmic tail (indicated by arrows in Fig. 1A) was cleaved most efficiently by caspase-3, and to a lesser extent by caspase-7, to produce an ϳ78-kDa product (indicated by circles) as well as a 17-kDa product (indicated by asterisks) (Fig. 1A). All other caspases tested failed to cleave the Dsg1 cytoplasmic tail.
The cytoplasmic tail of Dsg1 contains a caspase-3 consensus site (DXXD), DLRD, at amino acids 885-888 (Fig. 1C). Cleavage at Asp 888 would result in predicted products 78 and 17 kDa in size. To determine whether Dsg1 was cleaved by caspase-3 at this site, the obligate aspartic acid residue within this site (Asp 888 ) was changed to a similar glutamic acid (Glu) residue by site-directed mutagenesis. Both the wild type Dsg1 cytoplasmic tail and the mutated protein were then utilized as substrates for caspase-3 in an in vitro cleavage assay. The wild type Dsg1 tail was cleaved by caspase-3 to produce the 78-and 17-kDa products. In contrast, the mutant Dsg1 tail remained intact in the presence of caspase-3 (Fig. 1B), confirming that Dsg1 was specifically cleaved by caspase-3 at Asp 888 .
Apoptotic Stimuli Promote Caspase Cleavage of Ectopic Dsg1 in Epithelial Cells-As an initial approach to address whether Dsg1 is a potential physiological caspase substrate in apoptotic cells, we first generated stable, tetracycline-inducible A431 epidermoid vulvar carcinoma cell lines expressing full-length Dsg1, FLAG-tagged at the C terminus (Dsg1.FLAG). Because Dsg1 is not expressed in cultured cell lines under normal conditions, this manipulable system allowed us the opportunity to readily detect Dsg1 and its cleavage products in cells in which apoptosis is easily induced. Western blot analysis of A431/B4 whole cell lysates showed that Dsg1.FLAG was expressed at the expected size of about 160 kDa only in cells that had been treated with 4 g/ml Dox for 18 -24 h (Fig. 2A). The tightly regulated expression was also evident by immunofluorescence microscopy (Fig. 2B).
To investigate the processing of Dsg1 in cells undergoing apoptosis, A431/B4 cells ectopically expressing Dsg1 were exposed to UV radiation, a physiologically relevant environmental stimulus of the apoptotic pathway in keratinocytes (55). Cell lysates containing pooled floating and adherent cells were collected at time points from 0 -24 h later. Dsg1.FLAG was immunoprecipitated from the lysates with an antibody against the C-terminal FLAG tag and analyzed by Western blot to detect both full-length Dsg1 and C-terminal cleavage products. The transient appearance of a 17-kDa protein species, likely corresponding to the 17-kDa product observed in our in vitro studies, was detected concomitantly with a marked decrease in the total amount of immunoprecipitated full-length Dsg1 protein and a progressive increase in the total amount of apoptotic cells over time (Fig. 3A). Whereas the 17-kDa C-terminal Dsg1 cleavage product was detected from 3-12 h after UV treatment, it was most prominent 6 -9 h after induction of apoptosis. Pretreating the cells with the general caspase inhibitor ZVAD-fmk resulted in the maintenance of the full-length Dsg1 protein levels and prevented the appearance of the 17 kDa C-terminal cleavage product (Fig. 3A), indicating that the generation of this product was a caspasedependent event. The 17-kDa C-terminal cleavage product was also observed when normal human epidermal keratinocytes ectopically expressing Dsg1 were induced to undergo apoptosis by treatment with cycloheximide and tumor necrosis factor-␣-related apoptosis-inducing ligand (TRAIL) (data not shown), indicating that caspase-specific cleavage of the Dsg1 cytoplasmic tail is not cell type-specific nor apoptotic stimulus-dependent.
To address whether caspase-dependent proteolysis of Dsg1 is coordinated with the processing of other desmosome-associated proteins, we examined the temporal pattern of cleavage of the plaque proteins DP, Pg, and the associated intermediate filament protein keratin 18, in response to UV irradiation. Apoptotic cleavage of all of the proteins tested followed a similar time course (Fig. 3), which correlated with that observed for Dsg1 (Fig. 3A). Although cleavage of Pg began slightly earlier than that of other proteins examined (Fig. 3B), the principal cleavage products of DP, keratin 18, Pg, and Dsg 1, were most abundant 6 -12 h after UV treatment and had largely disappeared or become further processed by 24 h (Fig. 3, A-D).
To further map the products of Dsg1 processing that are generated during apoptosis, we examined cells for evidence of the N-terminal counterpart of the 17-kDa C-terminal product predicted to result from caspase-3 cleavage of Dsg1. At time points from 0 to 24 h after UV irradiation of Dsg1.FLAG-expressing A431/B4 cells, pooled adherent and floating cells were harvested, lysed, and subjected to immunoprecipitation with an antibody recognizing the N-terminal extracellular portion of Dsg1. As previously demonstrated in Fig. 3A, full-length Dsg1 was targeted for cleavage in cells exposed to UV radiation, which contributed to its progressive and caspase-dependent disappearance over time (Fig. 4A). In addition, a product of about 140 kDa was transiently detected in the cells that had undergone apoptosis but not in the control cells (Fig. 4A). The 140-kDa product was most abundant 6 -9 h after UV treatment, corresponding to the time course of the 17-kDa C-terminal product (Fig. 3A). The generation of the 140-kDa N-terminal product was inhibited when cells were pretreated with ZVAD-fmk before induction of apoptosis, indicating that this proteolytic event was caspase-dependent (Fig. 4A). Collectively, these data provide evidence that this 140-kDa product is the N-terminal counterpart to the 17-kDa C-terminal caspase-3 cleavage product of Dsg1.
Metalloproteinase Cleavage of the Dsg1 Extracellular Domain during Apoptosis Produces a 75-kDa Product-Because apoptotic proteolysis of other classical and desmosomal cadherins has been shown to involve shedding of an extracellular fragment into the culture media (35,36,38), we asked whether Dsg1 also undergoes extracellular cleavage during epithelial cell apoptosis. Culture media was collected over time from UV-irradiated Dsg1.FLAG-expressing A431/B4 cells and subjected to immunoprecipitation with an antibody against the extracellular domain of Dsg1. Immunoblot analysis revealed the progressive accumulation of a 75-kDa protein species in the media harvested 3-24 h after induction of apoptosis (Fig. 4B).
To characterize the enzyme(s) responsible for the generation of this 75-kDa product, cells were pretreated with ZVAD-fmk, MMP inhibitor 1 (MMPI-1) (shown to block matrix metalloproteinases 1, 3, 8, and 9 (56)), or TAPI-0 (a metalloproteinase inhibitor originally demonstrated to block a disintegrin and metalloproteinase (ADAM) family member, tumor necrosis factor-␣ convertase (57)) before UV irradiation. Generation of the 75-kDa extracellular Dsg1 fragment was completely blocked in the presence of TAPI-0 and inhibited in the presence of ZVAD-fmk but unaffected by MMPI-1 (Fig. 4C). The contribution of metalloproteinases to the apoptosis-induced extracellular cleavage of Dsg1 is similar to that observed for the classical cadherins, VE-cadherin and E-cadherin, but unlike that seen with other desmosomal cadherins (35,36,38). Taken together, our results suggest that optimal production of the 75-kDa extracellular cleavage product of Dsg1 requires caspase activity as well as metalloproteinase activity and is perhaps mediated by an ADAM family member.

Cell Surface Expression of Dsg1 Decreases Over Time during
Apoptosis-To investigate how the observed cleavage events affect the cell surface expression of Dsg1 during apoptosis, immunofluorescence microscopy was performed on control or UV-treated Dsg1.FLAG-expressing A431/B4 cells. The fixed cells were stained for Dsg1.FLAG with antibodies against the N-terminal extracellular domain, the C-terminal FLAG tag, and with DAPI to visualize the nuclear morphology. In the control cells with healthy nuclei, robust border staining of Dsg1 was evident as was strong co-localization of the extracellular and intracellular domains of Dsg1 at the cell periphery (Fig. 5). DAPI-stained apoptotic condensed, and fragmented nuclei (indicated by arrows) were observed 12 h after exposure to UV radiation. In apoptotic cells, a loss of the signal for both the extracellular and intracellular portions of Dsg1 was noted at cell borders, and a marked decrease in the peripheral co-localization of these signals was observed together with an accumulation of the Dsg1 C terminus in the cell cytoplasm (Fig. 5). This is likely due to intra-and extracellular cleavage of full-length Dsg1.FLAG, resulting in the removal and release of the fragments bearing epitopes recognized by these antibodies. Thus, our results suggest that the cell surface expression of Dsg1 in the A431/B4 cells decreases over time during UV-induced apoptosis.
Apoptotic Processing of Endogenous Dsg1 Resembles That of Ectopic Dsg1-The A431 cell system facilitated the initial identification and characterization of UV-induced Dsg1 processing, but to validate the physiological relevance of our findings we sought to investigate the apoptotic cleavage of endogenous Dsg1. For this purpose we utilized the spontaneously immortalized, non-tumorigenic adult human keratinocyte cell line, HaCaT. This line is accepted as a model for normal keratinocytes and is capable of reconstituting a differentiated epidermis when transplanted in vivo (48). Furthermore, HaCaT cells can be induced to express endogenous Dsg1 when cells are cultured under conditions that favor differentiation, and they undergo apoptosis in response to a UV stimulus (58), albeit at a higher dosage than that used for A431 cells. 5 To determine whether endogenous Dsg1 undergoes apoptotic proteolysis in the same manner as ectopically expressed Dsg1 in the A431 cells, HaCaT cells were cultured to induce Dsg1 expression and UVirradiated. Cell lysates were harvested 0 -72 h later and subjected to Western blot analysis. A time-dependent decrease in the levels of fulllength Dsg1 together with the generation of a 17-kDa product was observed (Fig. 6A). The generation of this 17-kDa product correlated with the activation of caspase-3 and an increase in the percentage of apoptotic cells. Activation of caspase-3, the increase in the percent of apoptotic cells, the decrease in full-length Dsg1, and the generation of the 17-kDa product were each inhibited by the general caspase inhibitor, ZVAD-fmk. Furthermore, a fragment of ϳ75 kDa was isolated  In addition, a progressive, caspase-dependent decrease in native Dsg1 was detected over time after UV treatment. This is demonstrated quantitatively by densitometry analysis of a less-exposed film, where numerical values represent amounts of full-length Dsg1 relative to that present in the control cells (bottom panel). B, the culture media was collected from Dsg1.FLAG-expressing A431/B4 cells 0 -24 h after exposure to UV radiation and subjected to immunoprecipitation with an antibody against the Dsg1 N terminus (982). The precipitated products were analyzed by Western blotting with an antibody that detects the N terminus of Dsg1 (P124). A 75-kDa specific cleavage product progressively accumulates over time after induction of apoptosis. C, A431/B4 cells expressing Dsg1.FLAG were pretreated with ZVAD-fmk (ZVAD), MMPI-1, or TAPI-0 (TAPI) for 1 h and then induced to undergo apoptosis with UV irradiation. After 8 h, the culture media was recovered from each sample and subjected to immunoprecipitation and Western blot as in panel B. The shedding of the 75-kDa Dsg1 product into the culture media is completely prevented in the presence of TAPI-0 and inhibited in the presence of ZVAD-fmk. D, schematic representation of the Dsg1 protein, which highlights the products resulting from the apoptotic processing of Dsg1. The relative location of the caspase-3 consensus site ( 885 DLRD 888 ) as well as the portions of the molecule that correspond to the caspase-mediated cleavage products (140-kDa product and 17-kDa product) and the predicted metalloproteinase generated product (75-kDa product) are from the culture media and increased in abundance over time after induction of apoptosis (Fig. 6B). The generation of this fragment was partially inhibited in the presence of ZVAD-fmk, indicating that caspases play a role in its generation. The extracellular shedding of the 75-kDa product was completely prevented in the presence of a panel of MMP inhibitors including TAPI-0, MMPI-1, and GM6001 (data not shown).
The effects of the apoptotic proteolysis of endogenous Dsg1 in UVtreated HaCaT cells were also examined by immunofluorescence microscopy. Using antibodies that allow us to detect specifically the N or C terminus of Dsg1, we observed that in control cells with normal nuclei, both the Dsg1 C terminus (Fig. 6C) and the N terminus (Fig. 6D) co-localized with DP at the cell-cell borders and were expressed in a uniform and organized fashion. UV-treated cells exhibited fragmented and condensed nuclei and a dramatic decrease in the organization and cell border localization of both the N terminus (Fig. 6D) and the C terminus (Fig. 6C) of Dsg1 together with DP. The cell border localization and organization of Dsg1 and DP could largely be restored in the presence of ZVAD-fmk, which also rescued the normal nuclear morphology (Fig. 6C). Taken together, these results demonstrate that the apoptotic targeting of Dsg1 by both caspases and MMP or ADAM family members is likely a physiologically relevant event.
Dsg1 Knockdown in Differentiated Keratinocytes Is Associated with Inhibition of Apoptosis-To investigate whether Dsg1 status is causally linked to programmed cell death, Dsg1 expression was suppressed in HaCaT cells by transduction of a Dsg1-specific shRNA retrovirus, which specifically knocks down Dsg1 but not Dsg2, -3, or -4 (data not shown). When the transduced cells were differentiated, we observed greater than 50% knockdown of Dsg1 expression in the Dsg1 shRNAexpressing cells (shDsg1) compared with the cells transduced to express a control GFP construct (Fig. 7A). Cells were UV-irradiated, lysates were harvested after 0 -72 h, and immunoblot analysis was performed to examine typical markers of apoptosis, such as caspase-3 and poly-(ADP-ribose) polymerase (PARP), both of which are efficiently and characteristically cleaved in apoptotic cells. As shown in Fig. 7A, PARP cleavage began earlier and ultimately went to completion in the GFPexpressing control cells, whereas in the shDsg1 cells, PARP was never totally processed during the time course of the assay. The time course and efficiency of caspase-3 cleavage and activation was similarly affected. Expression of full-length Dsg1 decreased over time in both cases due to previously demonstrated apoptotic targeting, whereas a 17-kDa Dsg1 product was observed after induction of apoptosis exclusively in the control GFP-expressing cells (data not shown).
To examine directly how Dsg1 expression affects epithelial cell death, HaCaT cells expressing a GFP control construct-only or both a Dsg1 shRNA and GFP construct (shDsg1/GFP) were differentiated then treated with UV irradiation. Cells were fixed after 0 -72 h and stained to detect Dsg1 and nuclear morphology by immunofluorescence microscopy (Fig. 7B). Dsg1 was detected in an organized fashion at cell-cell borders of control cells but was noticeably reduced in the shDsg1/GFP cells. Upon UV treatment, Dsg1 staining in control cells was diminished and less well organized than in the untreated cells, presumably due to intra-and extracellular processing of this cadherin. In addition, the number of apoptotic nuclei was increased in both UV-irradiated GFP control cells and shDsg1/GFP cells compared with their untreated counterparts. Interestingly, more control GFP cells appeared apoptotic at 24 h compared with the shDsg1/GFP cells. To determine the percentage of apoptotic cells in each sample, we counted apoptotic nuclei associated with robust Dsg1 and GFP expression (in GFP control cells) or reduced Dsg1 expression but steady GFP expression (in shDsg1/GFP cells) at each time point. At 24 h after UV irradiation, the percent cell death in the GFP control cells containing robust Dsg1 expression was ϳ6-fold higher than that of the shDsg1/GFP cells with suppressed Dsg1 expression (Fig. 7C). Dsg1 shRNA provided a protective effect even up to 48 h after UV treatment, when Dsg1 expression in control GFP cells is also decreased due to apoptotic targeting. Our results, when taken together, indicate a relationship between Dsg1 expression and susceptibility to UV-induced apoptosis.

DISCUSSION
Proteases target cadherin family proteins found in multiple adhesive junctions during apoptosis and cleavage of these substrates is thought to contribute to the changes in cell morphology and decreased cell-cell adhesion associated with programmed cell death (34 -36,38). The present report identifies a differentiation-specific desmosomal cadherin, Dsg1, as a novel caspase target processed during the dismantling of desmosomes in keratinocytes undergoing UV-mediated apoptosis. Specifically, we have shown that in UV-irradiated keratinocytes, ectopically and endogenously expressed Dsg1 undergo caspase-dependent cleavage, generating an N-terminal fragment of 140 kDa and a C-terminal fragment of 17 kDa. Our in vitro data suggest that caspase-3 is most likely primarily responsible for this cleavage event, although caspase-7, which recognizes the same DXXD consensus site as caspase-3 (59), was also capable of generating this C-terminal fragment, although less efficiently. Expression of caspase-1, -2, -3, -4, and -7 has been noted in cultured keratinocytes, and enzymes with characteristics of the caspase-3 subfamily (including caspases-3, -6, and -7) have been identified as components of human cornified skin extracts (60). Although the in vitro caspase cleavage of the Dsg1 cytoplasmic tail is modest, the cleavage in cells is quite robust. The enhanced cleavage observed in cultured cells may reflect conformational differences between the full-length Dsg1 protein in cells and the cytoplasmic tail construct in vitro and/or the action of multiple proteases (caspases and metalloproteinases) on Dsg1 in apoptotic cells.
We identified the consensus sequence, DLRD, at amino acids 885-888 of the full-length Dsg1 molecule as the site of cleavage by caspase-3 using mutational analysis. The N-and C-terminal products transiently produced upon caspase cleavage of Dsg1 at this site are generated in a caspase-dependent manner with a comparable time course and together, total the approximate size of the full-length protein. Because caspase-3 cleaves Dsg1 at Asp 888 , it is likely that the 140-kDa N-terminal product corresponds to Dsg1 1-888 , and the 17-kDa C-terminal product corresponds to Dsg1 889 -1049 . Although other desmosomal cadherin isoforms have been implicated as caspase substrates (36), the caspase-3 consensus site identified in Dsg1 is not conserved in Dsg2, Dsg3, or Dsg4. 5 The non-conserved nature of this site suggests that it may be FIGURE 6. Endogenously expressed Dsg1 in HaCaT cells is apoptotically processed in a similar fashion to ectopically expressed Dsg1 in A431 cells. A, HaCaT cells were differentiated to express endogenous Dsg1 and were treated with UV irradiation. Cell lysates were harvested after 0 -72 h and subjected to Western blot analysis. An antibody against the Dsg1 N terminus (P124) was used to detect full-length Dsg1. This antibody also detects a smaller, caspase-dependent band (indicated by the asterisk). Based on its molecular weight, this band could represent a fragment resulting from the extracellular cleavage of Dsg1 that remains adherent to the cell surface. The 17-kDa C-terminal cleavage product was detected with an antibody specific for the Dsg1 C terminus (4B2). An antibody against the full-length proform of caspase-3 was used to detect caspase-3 activation by its cleavage, indicated here by the loss of signal for the pro-form. Immunodetection of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used to indicate equal protein loading. The percentage of cells undergoing apoptosis at each time point was assessed by nuclear morphology analysis of parallel samples. Generation of the Dsg1 17-kDa product occurs over time concomitantly with the disappearance of the fulllength Dsg1, activation of caspase-3, and increase in the number of apoptotic cells. All are prevented in the presence of ZVAD-fmk (ZVAD). B, HaCaT cells endogenously expressing Dsg1 were induced to undergo apoptosis with UV irradiation. After 0 -72 h, the culture media was collected and subjected to immunoprecipitation using an antibody against the Dsg1 extracellular domain (982). The precipitated proteins were analyzed by Western blotting using an antibody specific for the Dsg1 N terminus (P124). The appearance of a 75-kDa product is progressively detected over time and inhibited in the presence of the general caspase inhibitor, ZVAD-fmk. C, control or UV-treated Dsg1-expressing HaCaT cells were analyzed for expression and localization of the Dsg1 C terminus and DP and evaluated for nuclear morphology using immunofluorescence microscopy. The Dsg1 C terminus was stained with 4B2 (red). DP was stained with NW6 (green). Nuclei were stained with DAPI (blue). Singly labeled images were combined to create a merged image. Pretreatment with ZVAD-fmk largely rescues the UV-induced changes. D, Dsg1-expressing HaCaT cells were untreated or induced to undergo apoptosis with UV irradiation. Seventy-two hours later cells were processed for immunofluorescence microscopy and stained for the N terminus of Dsg1 with 982 (green), for DP with NW6 (red), and with DAPI (blue) to visualize the nuclear morphology. UV irradiation results in an increase in apoptotic nuclei together with a decrease in the cell border localization and organization of both the Dsg1 C-and N terminus and DP. Similar results were obtained in multiple experiments. MW, molecular weight.
particularly important for unique functions of Dsg1. This site was mapped C-terminal to the region at which Pg interacts with Dsg1. However, cleavage at this site does not appear to affect the ability of Dsg1 to interact with its best characterized binding partner, Pg (data not shown). Nevertheless, because the function of the C-terminal RUD and desmoglein terminal domain motifs in Dsgs are currently unknown, potential interactions with as yet uncharacterized binding partners mediated through these unique regions could be compromised by caspase cleavage of Dsg1.
Upon induction of apoptosis, Dsg1 is not only targeted by intracellular caspases but also undergoes extracellular processing to produce a 75-kDa product in a manner that is inhibited by TAPI-0 but not by MMPI-1 in A431 cells, suggesting that Dsg1 may be processed by an MMP or ADAM family member. Interestingly, the inhibitor profile for this Dsg1 extracellular processing event in A431 cells more closely resembles that of the classical cadherins, E-cadherin and VE-cadherin (35,38), than the desmosomal cadherins Dsg3 and Dsc3, in which extracellular proteolysis can be inhibited by MMPI-1 but not TAPI-0 (36). Inhibition of the extracellular cleavage of endogenous Dsg1 in HaCaT cells was less selective since a panel of metalloproteinase inhibitors, including TAPI-0, MMPI-1, and GM6001, completely prevented this cleavage event equally well (data not shown). This was the primary FIGURE 7. Knockdown of Dsg1 expression is associated with an inhibition of UV-induced apoptosis in keratinocytes. A, HaCaT cells were retrovirally transduced to express a control GFP construct or a Dsg1 shRNA construct (shDsg1), then allowed to differentiate. Apoptosis was induced with UV irradiation, and after 0 -72 h cell lysates were harvested and subjected to Western blotting to detect the presence of Dsg1 and apoptosis markers. A specific antibody against Dsg1 (P124) was used to demonstrate knockdown of Dsg1 expression in the shDsg1 cells compared with the GFP control cells. GFP expression was detected with an antibody against GFP. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. Antibodies specific for PARP and full-length caspase-3 show apoptosis-associated cleavage of PARP from its 115-to its 85-kDa form as well as caspase-3 activation, indicated by the loss of signal for pro-caspase 3, respectively. Both occur earlier and more completely in the GFP cells compared with the shDsg1 cells. B, HaCaT cells were retrovirally transduced to express a GFP control construct or both a Dsg1 shRNA and a GFP construct (shDsg1/GFP). The cells were differentiated and then induced to undergo apoptosis with UV irradiation. At 0 -48 h cells were fixed and stained for immunofluorescence microscopy to detect Dsg1 (red) and the nuclear morphology with DAPI (blue). GFP epifluorescence is seen in green. Representative images taken at 24 h after UV treatment are shown and demonstrate that both the GFP cells and the shDsg1/GFP cells exhibit an increase in the number of apoptotic cells. C, the percent cell death at each time point was quantified by assessing the amount of apoptotic nuclei in GFP-positive cells that correlated with expression of Dsg1 in the GFP control cells or the lack of Dsg1 expression in the shDsg1/GFP cells at time points from 0 -48 h after UV treatment. Suppression of Dsg1 expression is associated with resistance to UV-induced apoptosis. Analysis of each sample was performed in triplicate, and data are presented as the mean Ϯ S.D. Asterisks represent a statistically significant result by Student's t test (p Ͻ 0. 05). These data are representative of results from multiple independent experiments. difference noted between the apoptotic processing of ectopic and endogenous Dsg1 and may be a result of biological differences between the two cell types. Interestingly, the generation of the 75-kDa extracellular Dsg1 product was consistently inhibited in the presence of the general caspase inhibitor, ZVAD-fmk, suggesting that caspase activity may be indirectly required for maximal efficiency of the metalloproteinase-mediated cleavage of the Dsg1 extracellular domain.
The extracellular domain of Dsg1 is thought to function in mediating calcium-dependent cell-cell adhesion (2), and cleavage of the Dsg1 extracellular domain by bacterial exfoliative toxins inhibits keratinocyte adhesion in the upper layers of the epidermis (61)(62)(63). Furthermore, previous reports suggest that extracellular cadherin fragments perturb intercellular adhesion (64,65). Thus, apoptotic cleavage of Dsg1 may not only render it incapable of mediating intercellular adhesion, but the extracellularly shed fragments of Dsg1 may interfere and compete for binding between full-length desmosomal cadherin molecules, thereby further contributing to the loss of cell-cell adhesion.
Several reports have recently highlighted the ability of classical cadherins to promote cell survival and to inhibit apoptosis (66 -68). However, we demonstrate for the first time that suppressing the differentiation-dependent Dsg1 in HaCaT keratinocytes decreases their susceptibility to UV-mediated apoptosis. These results suggest that, unlike classic cadherins, Dsg1 promotes apoptosis in keratinocytes subjected to genotoxic stress. Given the structural dissimilarity of the cytoplasmic domains of desmosomal cadherins and classic cadherins, it is perhaps not surprising that the desmosomal cadherins exhibit unique functional properties. Moreover, this provocative finding suggests that caspase cleavage of Dsg1 may play an active role in promoting or potentiating apoptosis in damaged epithelial cells.
UV radiation is a potent environmental hazard that can cause the appearance of individual sunburn cells in human skin and their accumulation in the superficial layers (Refs. 69 and 70; for review, see Ref. 71). Sunburn cells are identified by their pyknotic nuclei and shrunken cytoplasm (72) but undergo the morphological and biochemical changes typical of apoptotic cells, including loss of cell-cell adhesion and separation from neighboring cells. Sunburn cell formation and apoptosis in the epidermis is thought to be one mechanism to protect against the accumulation of DNA damage that results from exposure to UV radiation and other types of environmental stimuli. In contrast to the widespread expression of classical cadherins throughout the layers of the epidermis, Dsg1 expression is restricted to the upper epidermis where its regulated distribution may be important to ensure destruction of cells that have been damaged by UV irradiation. The apoptotic processing of Dsg1 along with the temporally coordinated targeting of other desmosome components likely contributes to the loss of these sunburned cells from the superficial epidermis. Future studies will attempt to determine whether it is the simple expression of Dsg1 or, rather, the apoptotic fragments generated by cleavage of Dsg1 under apoptotic conditions that mediate increased cell death. Indeed, the caspase-generated fragments of several proteins have been shown to promote apoptosis (28,73,74). Other molecules, when cleaved during apoptosis, acquire unique signaling functions (33). It will be interesting to discover if a similar scenario might occur after the processing of Dsg1.
Taken together our results indicate that UV irradiation-induced caspase and metalloproteinase cleavage of Dsg1 likely contributes to the coordinated dissolution of desmosome junctions and the disruption of intercellular adhesion during apoptosis of differentiated keratinocytes. In addition, our findings point to a previously unrecognized role for Dsg1 in the apoptotic destruction of keratinocytes damaged by UV exposure.