Caspase Proteolysis of the Integrin β4 Subunit Disrupts Hemidesmosome Assembly, Promotes Apoptosis, and Inhibits Cell Migration*

Caspases are a conserved family of cell death proteases that cleave intracellular substrates at Asp residues to modify their function and promote apoptosis. In this report we identify the integrin β4 subunit as a novel caspase substrate using an expression cloning strategy. Together with its α6 partner, α6β4 integrin anchors epithelial cells to the basement membrane at specialized adhesive structures known as hemidesmosomes and plays a critical role in diverse epithelial cell functions including cell survival and migration. We show that integrin β4 is cleaved by caspase-3 and -7 at a conserved Asp residue (Asp1109) in vitro and in epithelial cells undergoing apoptosis, resulting in the removal of most of its cytoplasmic tail. Caspase cleavage of integrin β4 produces two products, 1) a carboxyl-terminal product that is unstable and rapidly degraded by the proteasome and 2) an amino-terminal cleavage product (amino acids 1–1109) that is unable to assemble into mature hemidesmosomes. We also demonstrate that caspase cleavage of integrin β4 sensitizes epithelial cells to apoptosis and inhibits cell migration. Taken together, we have identified a previously unrecognized proteolytic truncation of integrin β4 generated by caspases that disrupts key structural and functional properties of epithelial cells and promotes apoptosis.

Apoptosis is a genetically regulated cellular suicide program activated by diverse signals that results in a series of stereotypical events culminating in cell death. These events include membrane blebbing, nuclear and DNA fragmentation, loss of cell adhesion, dismantling of the cytoskeleton, and packaging the nuclear and cytoplasmic remnants into apoptotic bodies which are engulfed and degraded by adjacent cells. Caspases are highly conserved cell death proteases that execute many of these stereotypical events by cleaving target proteins at aspartic acid residues (1). For example, caspases trigger the internucleosomal fragmentation of DNA by cleaving ICAD, an inhibitor of the caspase-activated DNase (CAD), thereby releasing ICAD from the inactive ICAD⅐CAD complex and activating CAD (2,3). Caspases dismantle the nuclear envelope and intermediate filament cytoskeleton by specifically proteolyzing the nuclear lamins, cytokeratins 14, 18, and 19 in epithelial cells, vimentin in mesenchymal cells, and desmin in muscle cells (4 -10). Caspases also disrupt the actin microfilament network by proteolyzing and activating the actin-severing protein gelsolin and by cleaving actin directly (11,12). In addition, caspases proteolyze several components of adherens junctions and desmosomes, intercellular junctions attached to the actin microfilament or cytokeratin intermediate filament networks, respectively, which play a critical role in cell-cell adhesion in epithelial tissues. Indeed, classical (E-cadherin) and desmosomal cadherins (desmoglein-1 and -3 and desmocollin-3), ␤-catenin, plakoglobin, and desmosomal plaque proteins (plakophilin-1, desmoplakin-1 and -2) are all caspase substrates (13)(14)(15)(16). Clearly, adherens junctions and desmosomes and their associated cytoskeletal networks have been extensively targeted for degradation by caspases. The proteolytic disassembly of these structures by caspases likely contributes to the disruption of cell-cell adhesion and the dramatic cytoskeletal reorganization that typifies apoptosis.
In an effort to systematically decipher the molecular mechanisms by which caspases induce apoptosis, we have used a small pool expression cloning approach to isolate caspase substrates (9,(17)(18)(19)(20). Here, we report the identification of the integrin ␤4 subunit as a novel caspase substrate. Integrins are a family of heterodimeric cell surface receptors composed of an ␣ and ␤ subunit that adhere cells to the extracellular matrix (ECM) and transmit signals from the ECM that regulate cell survival, proliferation, differentiation, and migration (21). The integrin ␤4 subunit is distinguished from other integrin subunits by virtue of several unique characteristics. In contrast to the short intracellular domains of other integrins, integrin ␤4 has a large cytoplasmic tail spanning ϳ1000 amino acids that consists of two pairs of fibronectin type III (FNIII) 2 repeats separated by a connecting sequence (22,23). Moreover, the integrin ␤4 subunit and its ␣6 partner (␣6␤4 integrin) are receptors for the extracellular matrix protein laminin-5, recently designated laminin-332 by a laminin nomenclature committee (24). ␣6␤4 integrin is organized into multiprotein cytoplasmic plaques called hemidesmosomes that anchor cells in the basal layer of epithelial tissues, including skin and the mammary gland, to the basement membrane. Hemidesmosomes are attached to the cytokeratin network, whereas other integrins are attached to actin microfilaments (22,23). Additional protein components of hemidesmosomes are BP180 (BPAG2), a type II transmembrane protein, BP230 (BPAG1), and the linker protein plectin. Importantly, the cytoplasmic tail of integrin ␤4 plays a key role in the localization of ␣6␤4 integrin to hemidesmosomal structures and in the recruitment of other hemidesmosomal proteins to these structures. Specifically, the first and/or second FNIII repeat and a portion of the connecting sequence are required for assembly of ␣6␤4 integrin into hemidesmosomes and plectin binding, whereas the third FNIII repeat and part of the connecting sequence mediate BP180 binding (25)(26)(27). BP230 interacts with the third and fourth FNIII repeats and the carboxyl terminus of integrin ␤4 (28). The paramount importance of hemidesmosomes in maintaining the structural integrity of certain epithelial tissues is evident from patients with junctional epidermolysis bullosa (JEB) with congenital pyloric atresia, a fatal skin blistering disease characterized by abnormal hemidesmosomes, detachment of epithelial cells from the basement membrane, and mutations in the ␤4 or ␣6 subunit genes (29,30). Indeed, mice with targeted deletion of the entire integrin ␤4 gene or the cytoplasmic domain lack hemidesmosomes and develop a JEB-like syndrome notable for increased detachment and death of epithelial cells (31)(32)(33). Moreover, deletion of integrin ␤4 or its carboxyl-terminal tail in keratinocytes results in profound abnormalities in cell migration and cell survival that likely impair wound healing (34,35). Taken together, these findings point to a critical role for the cytoplasmic tail of integrin ␤4 in the structural integrity, function, and survival of certain epithelial tissues.
In this report we demonstrate for the first time that the integrin ␤4 subunit is cleaved by caspase-3 and -7 at Asp 1109 in vitro and in cells undergoing apoptosis, thereby removing much of its cytoplasmic tail, including all four FNIII repeats. Caspase proteolysis of integrin ␤4 generates a carboxyl-terminal product that is unstable and rapidly degraded by the proteasome and an amino-terminal cleavage product that is unable to assemble into mature hemidesmosomes. In addition, caspase cleavage of integrin ␤4 sensitizes epithelial cells to apoptosis and inhibits cell migration. Collectively, our results point to a novel caspase-mediated truncation of integrin ␤4 that disrupts key structural and functional properties of epithelial cells and promotes apoptosis.
Cell Culture and Apoptosis Experiments-Immortalized, non-transformed human MCF-10A mammary epithelial cells (37) and human MDA-MB-435 breast cancer cells were purchased from the ATCC. MCF-10A cells were cultured in Dulbecco's modified Eagle's medium/F-12 medium (Invitrogen) supplemented with 5% horse serum (Invitrogen), 20 ng/ml epidermal growth factor (Sigma), 0.5 mg/ml hydrocortisone (Sigma), 100 ng/ml cholera toxin (Sigma), 10 g/ml insulin (Sigma), and penicillin/streptomycin (Invitrogen) as described (38). MDA-MB-435 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Immortalized integrin ␤4-deficient keratinocytes derived from a patient with JEB with pyloric atresia (34) were kindly provided by M. Peter Marinkovich, Stanford University School of Medicine. JEB keratinocytes were cultured in defined keratinocyteserum-free medium (Invitrogen) and penicillin/streptomycin (Invitrogen). For MCF-10A cells, apoptosis was induced by adding 1 g/ml recombinant tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), which was produced as described previously (39,40), or 1 M staurosporine (Sigma). To verify that the apoptotic cleavage of integrin ␤4 was mediated by caspases, cells were pretreated for 1 h with vehicle or 50 M zVAD-fmk, a broad spectrum caspase inhibitor, and then treated with TRAIL or staurosporine. In some experiments cells were pretreated for 1 h with vehicle or 100 nM epoxomicin (Calbiochem), a proteasome inhibitor, and then treated with TRAIL to determine whether the apoptotic cleavage product of integrin ␤4 was degraded by the proteasome. Apoptosis was induced in MDA-MB-435 breast cancer cells by adding 1 g/ml TRAIL in combination with 1 g/ml cycloheximide (Sigma); the latter agent sensitizes these cells to TRAIL-induced apoptosis. Apoptotic nuclei (fragmented or condensed) were scored by staining with 10 g/ml Hoescht 33258 (Sigma) as described previously (41). Two hundred cells per treatment condition were scored.
Transfection and Retroviral Transduction-MDA-MB-435 carcinoma cells were transiently transfected with 1 g of pcDNA3 plasmids containing full-length WT or mutant D1109E integrin ␤4 using Lipofectamine PLUS reagent (Invitrogen). For retroviral transduction of JEB keratinocytes, 10 g of pLEGFP-N1 vector (BD Biosciences) or pLEGFP-N1 plasmid containing WT integrin ␤4 (WT-␤4) or caspase-truncated integrin ␤4 (Tr-␤4) were transfected into the Phoenix amphotrophic retrovirus packaging cell line (ATCC), and the retroviral supernatant was prepared as described previously (42). JEB cells were incubated with retrovirus for 48 h, and pools were then generated by selection in 125 g/ml G418 (Invitrogen) for 10 days. The expression of proteins in these pools was confirmed by immunoblotting with a GFP antibody. Apoptosis was induced in JEB pools by treatment with 5 g/ml TRAIL, and apoptosis was scored by nuclear morphology as described under "Cell Culture and Apoptosis Experiments" in this section.
Cell Surface Expression of Integrin ␤4-JEB keratinocytes stably expressing GFP vector or GFP-tagged WT-␤4 or Tr-␤4 were collected, washed with PBS, and incubated with an integrin ␤4 extracellular domain antibody (3E1, Chemicon) for 45 min at RT. Cells were then washed, incubated with goat anti-mouse Cy5 secondary antibody for 30 min at RT, washed, and fixed in 0.5% paraformaldehyde for 5 min at RT. Cell surface expression of integrin ␤4 was determined by fluorescenceactivated cell sorting using a DakoCytomation CyAn Flow Cytometer.
Microscopy-JEB cells stably expressing integrin ␤4 constructs were cultured on glass coverslips, washed with PBS, fixed with 3.7% paraformaldehyde for 20 min at RT, and permeabilized with prechilled 0.5% Triton X-100 for 15 min. Slides were incubated with primary antibodies against plectin (43) (1:50 dilution) or BP180 (44) (1:50 dilution) for 1 h at 37°C. After washing, slides were incubated with secondary antibodies, Alexa Fluor 594 F(abЈ) 2 fragments of either goat anti-mouse or anti-rabbit IgGs (Molecular Probes) (1:200 dilution) for 1 h at 37°C. Slides were incubated with 0.5 ng/ml 4Ј,6-diamidino-2-phenylindole (Sigma) for 15 min at RT and then washed with PBS. Coverslips were mounted with ProLong Gold antifade reagent (Invitrogen). Confocal images of planes near the basal cellsubstrate surface were obtained with a Zeiss LSM510 UV META confocal microscope. Images were analyzed with Zeiss LSM5 software. For localization and colocalization studies, 100 cells in each of two experiments were scored.
Wound Closure Assay-JEB cells stably expressing GFP constructs were sorted by flow cytometry to ensure GFP expression, plated in 6-well plates, and grown overnight. The following day cells were washed with PBS, scraped with a pipette tip, washed again with PBS, and grown in fresh medium (medium was replenished every 24 h). Forty-eight hours later wounds were photographed using a Nikon ECLIPSE TE 2000-U microscope, and wound closure was measured with MetaMorph software (Version 6.1r6, Universal Imaging). For immunoblotting, confluent JEB cells stably expressing GFP-tagged WT-␤4 were scraped with a pipette tip in a grid-like pattern (multi-scratch is cleaved by caspase-3 into a ϳ58-kDa fragment (indicated by an arrow); this fragment was not observed when protein pool 59 was incubated with buffer control. 35 S-Labeled protein pools were generated from small pools of a human prostate adenocarcinoma cDNA library and incubated with 2.5 ng of recombinant caspase-3 as described (9,(17)(18)(19)(20). B, the cDNA encoding the ϳ75-kDa protein that is proteolyzed by caspase-3 into a ϳ58-kDa fragment was isolated from cDNA pool 59 by subdividing the pool and retesting smaller pools by the same approach. Clone 59A5 was sequenced and identified as a partial integrin ␤4 subunit cDNA. C, the full-length integrin ␤4 subunit is cleaved by caspase-3 in vitro. Full-length 35 S-labeled integrin ␤4 (denoted by an asterisk) is cleaved by 2.5 ng of caspase-3 into two products of ϳ130 and ϳ79 kDa (indicated by arrows).
wound assay) to increase the percentage of migrating cells for analysis.

Identification of the Integrin ␤4 Subunit as a Caspase-3 Substrate in Vitro by Small Pool Expression
Cloning-We have previously described an expression cloning method to systematically identify cDNAs encoding putative caspase substrates from small pools of a cDNA library (9,(17)(18)(19)(20). In these experiments caspase-3 was added to 35 S-labeled protein pools derived from small cDNA pools (48 cDNAs in each pool) of a human prostate adenocarcinoma cDNA library (Invitrogen). Pools with protein bands that were cleaved by caspase-3 were identified. An ϳ75-kDa protein (Fig. 1A, indicated by an asterisk) in 35 S-labeled pool 59 was proteolyzed by caspase-3 into a fragment of ϳ58 kDa (Fig. 1A, indicated by the arrow) in vitro. To identify the caspase-3 substrate corresponding to the ϳ75-kDa band in protein pool 59, we divided cDNA pool 59 into progressively smaller pools, 35 S-labeled these pools, and incubated them with caspase-3. In this way, a single cDNA (59A5) encoding an ϳ75-kDa protein (Fig. 1B, doublet, indicated by an asterisk) was cleaved by caspase-3 into a ϳ58-kDa product (Fig. 1B, doublet, indicated by an arrow) in vitro. Sequence analysis of cDNA 59A5 revealed that it encoded amino acids 593-1266 of the integrin ␤4 subunit (45). The cDNA contains two ATG codons near the 5Ј-end, which likely results in the observed doublet of the 35 S-labeled protein (Fig. 1B). Importantly, 35 S-labeled fulllength integrin ␤4 (Fig. 1C, indicated by the asterisk) was proteolyzed by caspase-3 into products of ϳ130 and ϳ79 kDa (Fig.  1C, indicated by arrows). Taken together, these results indicate that the integrin ␤4 subunit is proteolyzed by caspase-3 at a single site in vitro.
The Cytoplasmic Tail of the Integrin ␤4 Subunit Is Cleaved at Asp 1109 by Caspases-3 and -7 in Vitro-Because caspases are intracellular proteases that do not have access to the extracellular and transmembrane domains of integrin ␤4, we examined the sensitivity of the 35 S-labeled cytoplasmic tail of integrin ␤4 (amino acids 734 -1752) to a panel of recombinant caspases. Incubation of the 35 S-labeled ␤4 cytoplasmic tail with caspase-3 generated prominent fragments of ϳ79 and ϳ43 kDa ( Fig. 2A, indicated by arrows) and reduced the amount of the intact cytoplasmic tail ( Fig. 2A, denoted by an asterisk). Caspase-7, which cleaves substrates at a similar DEXD motif as caspase-3 (46), produced the same sized fragments but to a lesser extent. A faint band corresponding to the ϳ43-kDa product was also observed when the 35 S-labeled ␤4 cytoplasmic tail was incubated with caspase-8. However, caspases-1, -2, -5, -6, and -9 did not proteolyze the ␤4 cytoplasmic tail. Of note, the observed ϳ79-kDa product was similar in size to the fragment produced by caspase-3 cleavage of full-length 35 S-labeled ␤4 (Fig. 1C). These results indicate that the cytoplasmic tail of integrin ␤4 is preferentially proteolyzed by caspase-3 and -7 in vitro.
A potential caspase-3 and -7 cleavage motif (DEXD) (46) was identified in the ␤4 cytoplasmic tail (DELD 1109 2R) that would be expected to yield a ϳ79-kDa carboxyl-terminal cleavage product. To determine whether integrin ␤4 was indeed cleaved by caspase-3 and -7 at this site in vitro, we replaced Asp 1109 with a Glu residue and tested the sensitivity of the mutant D1109E ␤4 cytoplasmic tail to caspase proteolysis. Unlike the WT ␤4 cytoplasmic tail ( Fig. 2A), the 35 S-labeled D1109E ␤4 tail was not cleaved by caspase-3 or -7 (Fig. 2B), indicating that Asp 1109 is the bona fide caspase-3 and -7 cleavage site in vitro. A domain map of the integrin ␤4 protein reveals that caspase proteolysis at Asp 1109 removes much of the cytoplasmic tail, including all four fibronectin type III repeats and the connecting sequence (Fig. 2C). Taken together, these findings indicate that ␤4 integrin is cleaved by caspases-3 and -7 at Asp 1109 in its cytoplasmic tail, thereby removing several domains involved in hemidesmosome assembly and/or signaling.
Integrin ␤4 Is Proteolyzed at Asp 1109 by Caspases in Cells Undergoing Apoptosis and Produces an Unstable Cleavage Product That Is Degraded by the Proteasome-We next examined whether the integrin ␤4 subunit is cleaved in epithelial cells undergoing apoptosis. For these experiments we treated MCF-10A mammary epithelial cells, which express integrin ␤4 and assemble mature hemidesmosomes (47,48), with 1 g/ml TRAIL for 0 -8 h. The amount of full-length integrin ␤4  detected with an antibody directed against its carboxyl terminus decreased progressively over time, and there was no detectable full-length ␤4 8 h after TRAIL treatment (Fig. 3A, left  panel). Although no ␤4 cleavage product was observed (data not shown), the reduction in the amount of full-length ␤4 in response to TRAIL was completely suppressed by the broad spectrum caspase inhibitor zVAD-fmk, suggesting that caspases were responsible for the observed reduction in fulllength ␤4. The caspase substrate protein kinase C␦ was used as a positive control; it was cleaved into its expected size apoptotic fragment (49) within 2 h of TRAIL treatment, and its apoptotic proteolysis was blocked by zVAD-fmk. A reduction in fulllength ␤4 was also observed in MCF-10A cells treated with staurosporine, and this reduction was suppressed by zVADfmk (Fig. 3A, right panel). These data suggest that caspases cleave the integrin ␤4 subunit during apoptosis in vivo and produce an unstable proteolytic product. Indeed, caspase proteolysis of several proteins has been demonstrated to yield unstable cleavage products that are rapidly degraded by the proteasome (50 -53). Hence, we postulated that the integrin ␤4 carboxylterminal cleavage product was similarly degraded. Consistent with this hypothesis, a ϳ72-kDa ␤4 cleavage product was observed in TRAIL-treated MCF-10A cells preincubated for 1 h with 100 nM epoxomicin, a proteasome inhibitor (Fig. 3B). This proteolytic product is similar in size to the ϳ79-kDa 35 S-labeleded ␤4 cleavage product generated by caspase cleavage in vitro ( Fig. 2A). Importantly, both the disappearance of fulllength ␤4 and the appearance of the ϳ72-kDa cleavage product in MCF-10A cells treated with epoxomicin and TRAIL were inhibited by zVAD-fmk. These findings provide unequivocal evidence that caspases proteolyze the integrin ␤4 subunit into an unstable carboxyl-terminal cleavage product that is subsequently degraded by the proteasome in epithelial cells undergoing apoptosis.
To verify that integrin ␤4 is cleaved by caspases at Asp 1109 in apoptotic cells, we transiently transfected human MDA-MB-435 breast carcinoma cells, which lack endogenous integrin ␤4 subunit, with a full-length WT-␤4 or D1109E-␤4 cDNA (Fig. 3C). Twenty-four hours later cells were treated with 1 g/ml TRAIL and 1 g/ml cycloheximide to induce apoptosis (the latter drug sensitizes carcinoma cells to TRAIL). Full-length WT-␤4 (denoted by the asterisk) was rapidly cleaved into the ϳ72-kDa product; some cleavage was also evident at t ϭ 0 due to the toxicity of the transfection reagents (a prominent more rapidly migrating band that was also detected by the carboxyl-terminal integrin ␤4 Ab was also cleaved). Of note, in these transiently transfected cells, the ϳ72-kDa caspase cleavage product was detectable in the absence of proteasome inhibition, perhaps due to the high levels of the ectopically expressed ␤4. Conversely, D1109E-␤4 was resistant to proteolysis with little production of the ϳ72-kDa fragment in TRAIL-treated cells.
These results indicate that integrin ␤4 is cleaved by caspases at Asp 1109 in cells undergoing apoptosis.
Caspase Cleavage of Integrin ␤4 Disrupts Hemidesmosome Assembly-Because caspase cleavage of ␤4 integrin removes cytoplasmic domains required for its localization to hemidesmosomes and its binding to other hemidesmosomal proteins (25)(26)(27), we examined the ability of GFP-tagged full-length WT-␤4 and caspase-truncated ␤4 (amino acids 1-1109, Tr-␤4) to assemble into hemidesmosomes in integrin ␤4-deficient keratinocytes. These keratinocytes were derived from a patient with JEB and contain all of the protein components of hemidesmosomes except the integrin ␤4 subunit (34). Importantly, the addition of GFP to the carboxyl terminus of WT integrin ␤4 does not alter its ligand binding, localization to hemidesmosomes, or signaling (54 -56). JEB keratinocytes stably expressing GFP WT-␤4, GFP Tr-␤4, or control-GFP vector were gen-FIGURE 3. Integrin ␤4 is proteolyzed at Asp 1109 by caspases in cells undergoing apoptosis and produces an unstable cleavage product which is degraded by the ubiquitin-proteasome system. A, full-length integrin ␤4 is rapidly degraded in human MCF-10A mammary epithelial cells treated with TRAIL (top panel) by a caspase-dependent mechanism. MCF-10A cells were treated with 1 g/ml TRAIL or 1 M staurosporine for 0 -8 h. For the caspase inhibitor experiments, MCF-10A cells were preincubated for 1 h with 50 M zVAD-fmk, a pan-caspase inhibitor, and then treated with 1 g/ml TRAIL or 1 M staurosporine for 4 h. Immunoblotting was performed with antibodies recognizing the cytoplasmic tail of integrin ␤4, protein kinase C␦ (PKC␦) or tubulin. The percentage of apoptotic nuclei was determined in parallel experiments. B, the apoptotic integrin ␤4 caspase cleavage product is degraded by the proteasome. MCF-10A cells were preincubated with 20 M ZVAD-fmk, 100 nM epoxomicin (Epox), a proteasome inhibitor, or both for 1 h and then treated with 1 g/ml TRAIL (or untreated) for 6 h. C, mutant D1109E integrin ␤4 is not cleaved during apoptosis. MDA-MB-435 breast cancer cells (which lack integrin ␤4) were transiently transfected with either WT or mutant D1109E integrin ␤4 cDNA. Twenty-four hours after transfection cells were treated with 0.5 g/ml TRAIL and 1 g/ml cycloheximide. In B and C, full-length integrin ␤4 (asterisks) and the ϳ72-kDa cleavage product (arrow) are indicated. erated by retroviral infection and selection in G418. Stable expression of each construct was confirmed by immunoblotting, as was the absence of full-length ␤4 in JEB cells stably expressing Tr-␤4 or GFP vector (Fig. 4A). We next performed fluorescence-activated cell sorting analysis using an antibody that recognizes the extracellular domain of integrin ␤4. Both WT-␤4 and Tr-␤4 were expressed on the surface of retrovirally transduced JEB cells (Fig. 4B). Confocal microscopy images of planes near the cell-substrate surface revealed WT-␤4 in polarized basal structures characteristic of hemidesmosomal proteins in 72 Ϯ 4.9% of cells (Fig. 4C). In contrast, Tr-␤4 was often more diffusely distributed and was present in polarized basal structures in only 44 Ϯ 2.8% of cells. These results indicate that although Tr-␤4 is expressed on the cell surface, it often fails to incorporate into hemidesmosome-like structures.
Caspase Proteolysis of Integrin ␤4 Promotes Apoptosis-To determine whether caspase cleavage of integrin ␤4 alters the sensitivity of epithelial cells to apoptosis induction, we treated JEB pools stably expressing GFP vector or GFP-tagged WT-␤4, D1109E-␤4, or Tr-␤4 with 5 g/ml TRAIL for 0 -8 h. Both Tr-␤4 and WT-␤4, but not cleavage-resistant D1109E-␤4, sensitized JEB keratinocytes to TRAIL-induced apoptosis 2 and 4 h after treatment (Fig. 6). However, by 8 h the induction of apoptosis was similar in each of the JEB pools. WT-␤4 was cleaved by caspases as early as 2 h after TRAIL treatment (data not shown). These results indicate that caspase cleavage of integrin ␤4 sensitizes keratinocytes to TRAIL-induced apoptosis.
Caspase Cleavage of Integrin ␤4 Inhibits Keratinocyte Migration-Because caspase-3 has been implicated in cell migration in non-apoptotic epithelial cells (57), we examined whether integrin ␤4 was cleaved in migrating keratinocytes. To this end, confluent JEB cells stably expressing GFP-tagged WT-␤4 were scraped with a pipette tip in a grid-like pattern (multi-scratch wound assay). A reduction in the amount of fulllength WT-␤4 was observed at 48 h in this assay, and this reduction was inhibited by zVAD-fmk (Fig. 7A). These results indicate that integrin ␤4 is cleaved by caspases in migrating keratinocytes. To determine the effects of integrin ␤4 cleavage on cell migration, confluent JEB cells stably expressing GFP vector or GFP-tagged WT-␤4 or Tr-␤4 were scraped with a pipette tip, and wound closure was measured 48 h later. Although WT-␤4 cells were highly motile in this assay, Tr-␤4 cells exhibited little if any migration and were not significantly different from ␤4-deficient vector-transduced JEB cells (Fig. 7, B and C). Importantly, these migration differences were not due to differences in cell number (data not shown). These results suggest that caspase cleavage of integrin ␤4 inhibits keratinocyte migration.

DISCUSSION
We have demonstrated that the integrin ␤4 subunit is a novel caspase substrate that is cleaved by caspases-3 and -7 in vitro and in apoptotic epithelial cells. Caspases-3 and -7 proteolyze integrin ␤4 at a DEXD consensus caspase-3 and -7 cleavage motif (46), DELD 1109 2R, in its cytoplasmic tail. Interestingly, this aspartic acid residue and the entire cleavage motif in integrin ␤4 has been highly conserved during evolution (DELD-R in dogs and cattle and DETD-R in mice and rats), suggesting that caspase proteolysis of ␤4 may occur at this same site during apoptosis in other species as well. As illustrated in Fig. 2C, caspase cleavage of ␤4 removes much of its cytoplasmic tail, including all four of the FNIII repeats and the connecting sequence. The resulting carboxylterminal caspase cleavage product is unstable and is rapidly degraded by the proteasome. Indeed, the amino-terminal amino acid of this cleavage fragment (Arg 1110 ) is a destabilizing residue that likely targets this cleavage product for degradation by the ubiquitindependent N-end rule pathway (58). Caspase proteolysis of NF-B and the Drosophila inhibitor of apoptosis protein DIAP1 also generate unstable cleavage products that are degraded by the N-end rule pathway  . Caspase proteolysis of integrin ␤4 promotes apoptosis. JEB pools stably expressing GFP vector or GFP-tagged WT-␤4, D1109E-␤4, or Tr-␤4 were treated with 5 g/ml TRAIL for the indicated times, and apoptotic nuclei were scored as described under "Experimental Procedures." Data are presented as the mean Ϯ S.E. of three experiments (left panel). *, p Ͻ 0.05 versus D1109E; ***, p Ͻ 0.001 versus D1109E. The expression of each construct in JEB pools before TRAIL treatment was determined by immunoblotting using a GFP antibody (right panel). (51,52), suggesting that caspases and the ubiquitin-proteasome pathway may collaborate to efficiently degrade certain substrates during the execution of apoptosis.
We have also demonstrated that removal of the cytoplasmic tail of integrin ␤4 by caspases, resulting in a truncated ␤4 integrin (amino acids 1-1109), has important structural and functional consequences. Although arbitrary deletional analyses of the cytoplasmic tail of ␤4 have demonstrated its importance in hemidesmosome assembly and diverse functions (25-27, 33, 35), our results provide the first evidence that ␤4 indeed undergoes proteolytic removal of its cytoplasmic tail in vivo during a normal cellular process, namely, apoptosis. Regarding the structural consequences, caspase-truncated ␤4, unlike fulllength ␤4, is unable to assemble mature hemidesmosomal structures, which play a critical role in adhering epithelial cells to the basement membrane. Specifically, we have demonstrated that although Tr-␤4 is present on the cell surface, it is often diffusely distributed rather than localized in polarized basal structures characteristic of hemidesmosomes. We also observed that Tr-␤4 was completely impaired in its ability to recruit the hemidesmosomal protein plectin to hemidesmosome-like structures, whereas BP180 recruitment was partly impaired. One potential explanation for the more severe defect in plectin recruitment to hemidesmosomal structures by Tr-␤4 compared with BP180 is that BP180 is a membrane spanning protein that also interacts with ␣6 integrin and laminin-5 (59 -61). Hence, ␣6 integrin and laminin-5 may contribute to BP180 recruitment to hemidesmosomes even in the absence of the cytoplasmic tail of ␤4. In contrast, plectin is dependent on the first and/or second FNIII repeat and a portion of the connecting sequence of ␤4 (26,27), domains that are absent from Tr-␤4, for incorporation into hemidesmosomes. Collectively, our results suggest that caspase cleavage of integrin ␤4 disrupts mature hemidesmosomes, an event that may promote the detachment of epithelial cells from the basement membrane and subsequent extracellular matrix detachment-induced apoptosis or anoikis.
Consistent with this idea, we have shown that caspase cleavage of integrin ␤4 sensitizes keratinocytes to apoptosis. Specifically, we demonstrated that JEB cells stably expressing Tr-␤4 or WT-␤4 (which was cleaved in response to TRAIL treatment), but not caspase cleavage-resistant D1109E-␤4, were more sensitive to apoptosis at early time points after treatment with TRAIL. These findings are concordant with reports that deletion of various domains in the cytoplasmic tail of integrin ␤4 result in enhanced apoptosis in keratinocytes and mammary epithelial cells (35,62,63). Because the cytoplasmic tail of integrin ␤4 is required for laminin-5-induced activation of cell survival pathways such as NF-B and phosphatidylinositol 3-kinase (35,62,63), proteolytic removal of this domain by caspases might promote apoptosis by abrogating these survival signals.
In addition, we observed that integrin ␤4 is cleaved by caspases in migrating keratinocytes during wound healing, a finding that is consistent with a recent report implicating caspase-3 in cell migration of non-apoptotic epithelial cells (57). We have also demonstrated that proteolysis of integrin ␤4 has important functional consequences for cell migration. Although caspase-truncated ␤4 was unable to promote keratinocyte migration in a wound closure assay, full-length ␤4 robustly enhanced cell motility. These results indicate that caspase cleavage of integrin ␤4 inhibits cell migration. Our results are in agreement with previous reports demonstrating that integrin ␤4-deficient JEB keratinocytes identical to those used in these experiments or keratinocytes expressing a less extensively truncated ␤4 (amino acids 1-1355) are impaired in cell migration, an important component of wound healing (34,35). These defects in cell migration have been attributed to impaired Rac1, NF-B, and/or c-Jun NH 2 -terminal kinase activation in ␤4-deficient keratinocytes or keratinocytes expressing truncated ␤4. Indeed, it seems likely that caspase cleavage of integrin ␤4 alters its downstream signaling to these and other pathways by removing key signaling residues/domains, an hypothesis that will be systematically explored in future studies. Collectively, the findings presented here demonstrate a novel proteolytic mechanism that profoundly alters integrin ␤4 localization and function. Together with the recent demonstration that the intracellular domains of other receptors such as epidermal growth factor receptor and HER-2/ErbB2 are cleaved by caspases (64 -66), these findings suggest that FIGURE 7. Caspase cleavage of integrin ␤4 inhibits keratinocyte migration. A, immunoblot analysis of JEB cells stably expressing GFP-tagged WT-␤4 that were subjected to a multi-scratch wound assay as described under "Experimental Procedures" in the absence or presence of 50 M zVAD-fmk. Lysates were prepared 0 -48 h after scratching. B, photomicrograph of a wound closure experiment. Confluent JEB cells stably expressing GFP vector or GFP-tagged WT-␤4 or Tr-␤4 were scraped with a pipette tip, and wound closure was assessed at 48 h as described under "Experimental Procedures." C, data are presented as the mean Ϯ S.E. of three experiments. *, p Ͻ 0.05 versus vector control.