The cytoplasmic domain of the integrin β4 subunit mediates both association with the hemidesmosomal cytoskeleton and recruitment of the signaling adaptor protein Shc. To examine the significance of these interactions during development, we have generated mice carrying a targeted deletion of the β4 cytoplasmic domain. Analysis of homozygous mutant mice indicates that the tail‐less α6β4 binds efficiently to laminin 5, but is unable to integrate with the cytoskeleton. Accordingly, these mice display extensive epidermal detachment at birth and die immmediately thereafter from a syndrome resembling the human disease junctional epidermolysis bullosa with pyloric atresia (PA‐JEB). In addition, we find a significant proliferative defect. Specifically, the number of precursor cells in the intestinal epithelium, which remains adherent to the basement membrane, and in intact areas of the skin is reduced, and post‐mitotic enterocytes display increased levels of the cyclin‐dependent kinase inhibitor p27Kip. These findings indicate that the interactions mediated by the β4 tail are crucial for stable adhesion of stratified epithelia to the basement membrane and for proper cell‐cycle control in the proliferative compartments of both stratified and simple epithelia.
Basement membranes are thin, continuous sheets of specialized extracellular matrix (ECM), which support epithelial and other cells and separate them from the underlying interstitial connective tissue (Yurchenco and O'Rear, 1994; Timpl and Brown, 1996). In addition to promoting cell adhesion and cytoskeletal organization, basement membranes influence the proliferation and differentiation of cells. The effects of basement membranes on cellular behavior are likely to be mediated by integrins (Hynes, 1987; Ruoslahti and Pierschbacher, 1987; Hemler, 1990). Upon binding to ECM ligand, integrins cluster on the plasma membrane and interact with the cytoskeleton, thereby promoting the assembly of adhesive junctions, such as focal adhesions and hemidesmosomes (Borradori and Sonnenberg, 1996; Burridge and ChrzanowskaWonidka, 1996). In addition, integrins are coupled to intracellular signaling pathways potentially able to regulate the cell cycle (Giancotti, 1997). In particular, a class of integrins, which include the α6β4 laminin receptor, the α1β1 collagen IV/laminin receptor, the α5β1 fibronectin receptor and the broad specificity receptor αvβ3, are linked to the Ras‐ERK signaling pathway and control the cell cycle by the signaling adaptor protein Shc (Mainiero et al., 1995, 1997; Wary et al., 1996).
The integrin α6β4 is a receptor for various isoforms of the basement membrane component laminin and binds with the highest apparent affinity to laminin 5 (Giancotti, 1996). In vivo, α6β4 is expressed in a number of tissues and cell types that are in contact with a basement membrane, including both simple and stratified epithelia (Kajiji et al., 1989), Schwann cells (Sonnenberg et al., 1990; Einheber et al., 1993), and a subset of endothelial cells (Kennel et al., 1992; Klein et al., 1993) and thymocytes (Wadsworth et al., 1992). In contrast to all the other known α and β subunit cytoplasmic domains, which are relatively short, the intracellular portion of the β4 subunit measures over 1000 amino acids in length and contains, in its C‐terminal half, two pairs of type III fibronectin‐like repeats separated by a 142 amino acid connecting segment (Hogervorst et al., 1990; Suzuki and Naitoh, 1990). While β1 and αv subunit‐containing integrins interact with the actin cytoskeleton and localize to focal adhesions, α6β4 is found concentrated at hemidesmosomes both in cultured cells and in vivo (Carter et al., 1990; Stepp et al., 1990), suggesting that the β4 tail specifies association with the hemidesmosomal cytoskeleton.
Hemidesmosomes are complex adhesive junctions that link the basement membrane to the intracellular keratin cytoskeleton and are found exclusively in the basal cell layer of stratified and transitional epithelia (Borradori and Sonnenberg, 1996). Gene transfer studies in cultured cells have provided evidence that the unique cytoplasmic domain of β4, and specifically a region comprising the first pair of type‐III fibronectin‐like modules and the connecting segment, is required for association of α6β4 with the hemidesmosomal cytoskeleton (Spinardi et al., 1993). Subsequent studies have indicated that ligation of α6β4 activates an integrin‐associated kinase and causes phosphorylation of the β4 cytoplasmic domain at multiple tyrosine residues (Mainiero et al., 1995). Phosphorylation of a tyrosine activation motif (TAM) located in the connecting segment is likely to be required for association with the hemidesmosomal cytoskeleton, because mutations at either one of the two tyrosine residues in the TAM abolish the incorporation of α6β4 in hemidesmosomes (Mainiero et al., 1995). The ability of a truncated β4 subunit to exert a dominant‐negative effect on hemidesmosome assembly without inhibiting initial adhesion to laminin 5 (Spinardi et al., 1995) and the absence of hemidesmosomes in the skin of α6 and β4 knock‐out mice (Dowling et al., 1996; Georges‐Labouesse et al., 1996; van der Neut et al., 1996) have indicated that α6β4 plays a crucial role in the assembly of hemidesmosomes and their linkage to the keratin filament system. Although it is widely assumed that integrin interaction with the cytoskeleton at adhesive junctions such as hemidesmosomes is necessary to consolidate cell adhesion (Alberts et al., 1994), this hypothesis has not yet been tested directly in vitro or in vivo. For example, although mice carrying null mutations at the β4 or α6 locus die a few hours after birth from a severe form of epidermal blistering (Dowling et al., 1996; Georges‐Labouesse et al., 1996; van der Neut et al., 1996), it is not clear whether their cell adhesion defect is caused by the absence of α6β4‐mediated binding to laminin 5, the lack of integrin association with the cytoskeleton, or both.
Several observations suggest that ligation of α6β4 may also promote cell proliferation. In the skin and other stratified squamous epithelia, the expression of α6β4 is restricted to the actively proliferating basal keratinocytes (Kajiji et al., 1989). When these cells detach from the laminin‐rich basement membrane to migrate to the upper epidermal layers, they exit from the cell cycle and begin to differentiate (Hall and Watt, 1989). Similarly, depriving cultured keratinocytes of anchorage to their ECM, which is rich in laminin 5 (Carter et al., 1991; Rousselle et al., 1991), results in withdrawal from the cell cycle and differentiation (Green, 1977). Finally, squamous carcinoma cells endowed with high proliferative potential often express elevated levels of α6β4 (Kimmel and Carey, 1986; Wolf et al., 1990), and overexpression of the integrin augments the invasive potential of breast carcinoma cells (Shaw et al., 1997). These results suggest that α6β4 can facilitate both invasion and growth in tumor cells.
In accordance with these observations, recent biochemical studies have indicated that α6β4 activates signaling pathways able to influence cell proliferation. In particular, ligation of α6β4 results in recruitment of the adaptor protein Shc. Upon association with α6β4, Shc is phosphorylated on tyrosine and thereby combines with the Grb2/mSOS complex (Mainiero et al., 1995). This process results in the activation of both the Ras‐ERK and the Rac‐JNK mitogen‐activated protein (MAP) kinase pathways (Mainiero et al., 1997). The α6β4 integrin has also been shown to activate phosphatidylinositol 3‐OH kinase (PI‐3K) (Shaw et al., 1997). Since inhibition of PI‐3K blocks the activation of JNK by α6β4, it is likely that this lipid kinase couples α6β4‐mediated activation of Ras to the Rac‐JNK pathway (Mainiero et al., 1997). In cultured keratinocytes, adhesion to laminin 5 mediated by α6β4 induces transcription from the Fos serum response element (SRE) and promotes progression through the G1 phase of the cell cycle in the presence of epidermal growth factor (EGF). In contrast, adhesion mediated by α2β1, which is not linked to Shc signaling, results in exit from the cell cycle even in the presence of otherwise mitogenic concentrations of EGF (Mainiero et al., 1997). These findings suggest that the signals from α6β4 and the EGF receptor converge on the Ras‐ERK pathway, ultimately resulting in transcription of immediate‐early genes and progression through G1. Although it is clear that α6β4 can promote cell proliferation under defined conditions in vitro, it remains to be established whether α6β4 signaling influences cell proliferation in vivo.
In order to examine the biological significance of the intracellular functions of α6β4 in vivo, we have generated mice carrying a targeted deletion of the β4 cytoplasmic domain. Analysis of these mice indicates that the β4 cytoplasmic domain is necessary for stable adhesion of stratified epithelia to the basement membrane and for proper cell cycle control in both simple and stratified epithelia.
Results and discussion
Targeted deletion of the β4 cytoplasmic domain
Homologous recombination in embryonic stem (ES) cells was used to introduce a stop codon in the mouse β4 gene immediately after the sequences encoding the transmembrane domain. As illustrated in Figure 1A, the replacement vector consisted of a fragment of ∼7 kb of the mouse β4 gene interrupted immediately after the exon encoding the transmembrane segment by an insertion cassette containing a stop codon followed by an SV40 polyadenylation site and a neomycin resistance gene. Homologous recombination at the wild‐type locus was expected to generate a mutant locus encoding a β4 subunit truncated after Lys734, the halt‐transfer stop signal. Gene transfer experiments in cultured cells have indicated that this tail‐less β4 subunit combines efficiently with α6 and is exported regularly to the cell surface, but is unable to associate with Shc (Mainiero et al., 1997) or the hemidesmosomal cytoskeleton (data not shown). Southern blot analysis indicated successful incorporation of the desired mutation in ∼10% of the doubly selected ES cell clones (Figure 1B and data not shown). PCR analysis on the intercrosses between heterozygous mice carrying the targeted deletion revealed that the mutation was transmitted with the expected Mendelian frequency (Figure 1C). While heterozygous mice appeared completely normal, all homozygous mutant animals presented with extensive blistering of the skin, particularly in the abdomen, the ventral portion of the torso, the paws and the tip of the tail (Figure 1D). They were unable to feed, and died within a few hours after birth, presumably from dehydration. To minimize damage to the skin, all subsequent analyses were conducted on mice delivered by Cesarean section by embryonic day (E) 18.5.
To examine if the tail‐less β4 subunit encoded by the mutant allele was expressed correctly, we subjected the skin of wild‐type and mutant mice to Northern blotting and immunoprecipitation analysis. As shown in Figure 2A, the recombinant truncated mRNA had the expected size, but was significantly more abundant than the wild‐type mRNA, presumably as a result of increased stability and/or transcription. If the biosynthesis of β4 exceeds that of α6, the fraction of unpaired β4 is degraded rapidly in the endoplasmic reticulum (ER) (Giancotti et al., 1992). In accordance with this finding, immunoprecipitation experiments indicated that the amount of tail‐less β4 associated with α6 in homozygous mutant mice was comparable with that of wild‐type β4 paired with α6 in wild‐type mice (Figure 2B). To verify that the levels of tail‐less α6β4 at the cell surface were comparable with those of wild‐type integrin, we performed fluorescence‐activated cell sorting (FACS) analysis on CD25+ thymocytes derived from homozygous mutant and wild‐type mice. These cells express homogeneous levels of α6β4 at their surface (Wadsworth et al., 1992) and can be analyzed easily without subculturing. The results demonstrated that the level of expression of tail‐less α6β4 at the cell surface was similar to that of wild‐type integrin (Figure 2C). Similar results were obtained with primary keratinocytes (not shown). We concluded that the tail‐less integrin encoded by the mutant allele was expressed correctly at the cell surface.
Characterization of the cell adhesion defect
Histological analysis of the skin of mutant mice showed wide areas of separation of the epidermis from the dermis (Figure 3A). In several instances, the epidermis remained connected to the dermis by short filaments emanating from the basal keratinocytes. Cross‐sections at the pyloric junction revealed that the stratified epithelium of the esophagus had detached from the underlying connective tissue and degenerated (Figure 3A). Similar signs of degeneration were also observed in cross‐sections of the largest (and presumably oldest) epidermal blisters. No signs of apoptosis in either affected or unaffected areas of the skin were revealed by TdT‐mediated dUTP‐biotin nick end labeling (TUNEL) analysis (data not shown), and electron microscopy indicated that cell death in the oldest blisters was due to necrosis, presumably from the interrruption of nutrient exchange with the dermal interstitium (Figure 6, bottom left panel). These results are consistent with the observation that primary keratinocytes denied anchorage to the ECM exit from the cell cycle and begin to differentiate rather than undergoing apoptosis (Gandarillas et al., 1997). The histological abnormalities of tail‐less β4 mice resembled those observed in the human disease junctional epidermolysis bullosa with pyloric atresia (PA‐JEB), which is caused by mutations in the α6 or β4 gene (Vidal et al., 1995; Ruzzi et al., 1997), as well as in α6 and β4 knock‐out mice (Dowling et al., 1996; Georges‐Labouesse et al., 1996; van der Neut et al., 1996). In contrast to the epidermis, the intestinal epithelium of homozygous mutant mice did not detach from the underlying mesenchyme (Figure 3A), suggesting that the cytoplasmic domain of β4 is essential for stable adhesion of stratified, but not simple, epithelia. No signs of apoptosis were detected in either wild‐type or mutant intestinal epithelium at E18.5 (data not shown). Thus, signaling by the cytoplasmic domain of β4 may contribute to, but does not appear to be required for, epithelial survival.
We next examined the mechanism of epidermal detachment in mutant mice. Immunofluorescent staining of unaffected areas of the skin indicated that the tail‐less α6β4 was concentrated normally at the basement membrane junction. This suggests that the tail‐less integrin binds efficiently to laminin 5, and this binding is sufficient for recruitment to the basal surface in the absence of cytoskeletal interactions (Figure 3B). The immunostaining for laminin 5 (as well as collagen IV and laminin 1, data not shown) in mutant mice was linear along the basement membrane and had the same intensity as that observed in normal mice. The pattern of expression of other epidermal integrins, including α1β1, α2β1 and the basement membrane assembly receptor α3β1 (DiPersio et al., 1997), was also unchanged in the skin of mutant mice (data not shown). Finally, the ultrastructure of basement membrane in the skin of mutant mice appeared normal at the electron microscopic level (Figure 6, top right). These observations suggest that the mutation in β4 does not affect the expression of other integrins or the assembly of the basement membrane.
Immunofluorescent analysis of detached areas of the skin of mutant mice indicated that the anti‐laminin 5 staining demarcated the floor of the lesions, indicating that the skin had split above the basement membrane. However, a significant fraction of tail‐less α6β4 remained attached to the basement membrane, while keratin 5 was confined entirely to the epidermis above the roof of the blister (Figure 3B). In accordance with this finding, electron microscopic analysis indicated that remnants of basal cell projections remained attached to the basement membrane (Figure 6, bottom right panel). The observation that a significant fraction of tail‐less α6β4 appears to be pulled out of the membrane of basal keratinocytes when these cells detached from the basement membrane is consistent with the finding that the tail‐less integrin binds efficiently to laminin 5, but cannot integrate with the cytoskeleton.
To investigate further the adhesive function of tail‐less α6β4, we performed adhesion assays on primary keratinocytes derived from wild‐type and homozygous mutant mice. The cells were plated on microtiter wells coated with different concentrations of laminin 5 in the presence of a monoclonal antibody that blocks the adhesive function of all β1 integrins, including α3β1, the only other known laminin 5 receptor in keratinocytes. As shown in Figure 4A, the keratinocytes expressing tail‐less α6β4 adhered to laminin 5 as well as cells expressing the wild‐type integrin in an adhesion assay. However, the results of a laminar flow detachment assay revealed that the adhesive strength mediated by tail‐less α6β4 was significantly lower than that of the wild‐type integrin (Figure 4B). These results indicate that while the tail‐less integrin binds efficiently to laminin 5, it is unable to mediate stable adhesion.
To examine the role of the β4 tail in hemidesmosome assembly, non‐lesional areas of the skin of mutant mice were subjected to immunofluorescence and electron microscopic analysis. Immunofluorescent staining indicated that the cytoskeletal component HD1/plectin, which interacts with the β4 cytoplasmic domain (Niessen et al., 1997; Reznieczek et al., 1998), was distributed diffusely in the cytoplasm of basal keratinocytes instead of being concentrated at the basement membrane junction (Figure 5). The putative adhesion receptor bullous pemphigoid antigen 2 (BPAG2), which is thought to be recruited to hemidesmosomes in response to a signal generated by the β4 cytoplasmic domain (Borradori et al., 1997), was partially polarized at the basal surface of keratinocytes in both wild‐type and mutant mice, suggesting that the ability of BPAG2 to bind to an extracellular ligand may be sufficient for targeting at the basement membrane junction (Figure 5). The hemidesmosomal plaque component BPAG1, which is involved in interaction with the keratin cytoskeleton (Guo et al., 1995), was concentrated at the basement membrane junction despite the absence of the β4 tail (Figure 5), perhaps because it can interact at least indirectly with BPAG 2. These observations indicate that the cytoplasmic domain of β4 is necessary for the recruitment of HD1/plectin, but not BPAG2 and BPAG1, to the basement membrane zone. Finally, electron microscopic analysis revealed that the epidermis of mutant mice completely lacked hemidesmosomes (Figure 6, top right). The desmosomes were instead intact (Figure 6, top right, insert). Taken together, these results demonstrate that the β4 cytoplasmic domain is necessary for nucleation of hemidesmosomes and indicate that the adhesive defect of mutant mice is caused by the inability of tail‐less α6β4 to mediate this function.
Analysis of the cell signaling defect
In the next series of experiments, we analyzed if α6β4 signaling regulated cell proliferation in stratified and simple epithelia. In the adult epidermis, the expression of α6β4 is restricted to the basal layer, which contains cells with proliferative capacity (Giancotti, 1996). To examine the role of α6β4 signaling in epidermal proliferation, skin sections from wild‐type and mutant mice at E18.5 were stained with monoclonal antibody (mAb) Ki‐67, which is directed toward a nuclear protein expressed exclusively during the S, G2 and M phase of the cell cycle (Schlütter et al., 1993). To rule out effects caused by disrupted adhesion, we examined only non‐lesional areas of mutant skin. As shown in Figure 7A, the epidermis of mutant mice contained significantly fewer Ki‐67‐positive cells per linear millimeter of basement membrane (33.7 ± 4.0) than that of wild‐type mice (61.7 ± 8.1). In addition, in normal mice at E18.5, both the basal and immediately suprabasal layers contained a large fraction of Ki‐67‐positive cells, but in mutant mice cell proliferation was largely restricted to the basal layer (Figure 7A). This result indicates that the cytoplasmic domain of β4 is required for optimal proliferation of epidermal precursor cells.
The integrin α6β4 is also expressed in several simple epithelia including the gastrointestinal tract (Simon‐Assmann et al., 1994). The epithelium of the intestine is renewed continually from a small number of proliferating cells concentrated in invaginated structures named crypts. Proliferation ceases in the upper segment of the crypt, and post‐mitotic enterocytes migrate toward the tip of the villus and begin to differentiate (Gordon and Hermiston, 1994). Different laminin variants are expressed along the crypt–villus axis: laminin 2 is enriched in the crypts, laminin 1 is more abundant in villi and laminin 5 is expressed in both (Simon‐Assmann et al., 1994; Orian‐Rousseau et al., 1996). Laminin 1 has been shown recently to contribute to the establishment and maintenance of differentiation in enterocyte‐like CaCo2 cells (De Arcangelis et al., 1996). The role of other laminin isoforms and their integrin receptors in intestinal function remains to be determined.
To determine if α6β4 signaling also regulated proliferation in simple epithelia, we focused on the intestine. In the proximal duodenum at E18.5, α6β4 and laminin 5 were concentrated at the basement membrane junction in both intervillar spaces and villi (data not shown), and cell proliferation was confined to the intervillar spaces, which invaginate later to form the crypts (Figure 7A). We observed that the total number of Ki‐67‐positive cells per intervillar space was lower in mutant mice (8.2 ± 1.4) than in wild‐type mice (12.9 ± 3.0). In addition, while in normal mice at this stage of development the basal portion of most villi contained a large fraction of Ki‐67‐positive enterocytes, cell proliferation was largely restricted to the intervillar spaces in mutant mice. Since adhesion in these sections was intact, these findings indicate that signals transduced by the β4 tail are necessary to maintain a normal proliferative compartment in both the epidermis and intestine.
Previous studies have indicated that cells denied adhesion to the ECM, and thereby arrested in mid‐G1, contain elevated levels of cyclin‐dependent kinase (cdk) inhibitors such as p27Kip (Fang et al., 1996; Zhu et al., 1996). We therefore stained sections of duodenum from wild‐type and mutant mice with antibodies to p27Kip. While in normal mice the enterocytes of intervillar spaces and the basal portion of villi did not contain detectable levels of p27Kip, in mutant mice the immunostaining for p27Kip extended to the base of villi with frequent p27Kip‐positive cells in the intervillar spaces (Figure 7B, arrows). Accordingly, the intervillar spaces of mutant mice contained fewer p27‐negative cells (9.7 ± 2.6) than those of wild‐type mice (15.6 ± 4.2). In addition, the nuclei of post‐mitotic enterocytes in mutant mice were stained more intensely by anti‐p27 antibodies than those in wild‐type mice. Although the suprabasal layers of skin contain another cdk inhibitor, p21Cip, we did not detect increased accumulation of this inhibitor in the suprabasal layers of mutant mice (data not shown). These results suggest that α6β4 signaling contributes to the regulation of p27Kip, but not p21Cip. Taken together, the reduced Ki‐67 labeling and increased p27Kip staining in the intestine of mutant mice are consistent with the conclusion that α6β4 signaling is necessary for optimal proliferation in vivo.
Finally, we examined whether the reduction of progenitor cells associated with the tail‐less B4 defect affected differentiation in the epidermis or intestine. Immunofluorescent analysis indicated that the pattern of expression of epidermal differentiation markers involucrin (data not shown), filaggrin and loricrin was similar in wild‐type and mutant mice at E18.5 (Figure 8). In addition, assays of enzymatic activity in situ indicated that the level of expression and localization of lactase in the intestine were similar in wild‐type and mutant mice at E18.5 (Figure 8). These results, together with those of histological analysis, suggest that α6β4 signaling is not required for normal differentiation of skin and intestine during embryogenesis.
It is well established that normal cells must adhere to the ECM in order to proliferate in vitro (Giancotti and Mainiero, 1994). The phenotype of mice lacking the β4 tail provides direct evidence that integrin signaling regulates the cell cycle in vivo. In particular, the observation that the number of proliferating cells in the basal layer of epidermis and intervillar space of intestine is significantly reduced in mutant mice, despite the presence of other Shc‐linked integrins in these cells, suggests that the combined input from multiple Shc‐linked integrins is required for optimal proliferation. Previous studies have indicated that the basal layer of human epidermis contains two types of precursor cells that can be identified on the basis of their β1 integrin expression levels: the slow‐cycling stem cells, which express high levels of α3β1 and α2β1; and the more rapidly proliferating transit amplifying cells, which display lower levels of β1 integrins (Jones and Watt, 1993; Jones et al., 1995). Notably, α6β4 is expressed at similar levels on both stem cells and transit‐amplifying cells (Jones and Watt, 1993; Jones et al., 1995), and may thus affect the proliferation of both types of precursor cells.
In accordance with the observation that anchorage to the ECM is required for proper down‐regulation of the cdk inhibitor p27Kip in early‐to‐mid G1 (Zhu et al., 1996), we have found that in tail‐less β4 mice the enterocytes located at the base of villi display increased nuclear levels of p27Kip. In contrast, the post‐mitotic keratinocytes did not contain increased levels of the other cdk inhibitor, p21Cip, expressed in the epidermis. These observations suggest that increased levels of p27 Kip in the intestinal precursor cells of mutant mice may contribute to the premature withdrawal from the cell cycle. However, the presence of a significant proliferation defect in the basal layer of the epidermis suggests that the down‐regulation of cdk inhibitors is not the only mechanism by which integrin signaling promotes cell cycle progression. Other mechanisms, such as the ability of integrins and growth factor receptors to coordinately regulate the expression of immediate‐early genes, including D‐type cyclins, are likely to play a crucial role.
The proliferation defect observed in mice lacking the β4 tail does not result in a reduced production of differentiated cells in the skin and intestine of E18.5 embryos. This is not surprising because the precursor cell compartment in normal skin and intestine at this stage of development is much larger than in the adult, and thus presumably exceeds the physiological requirements of the tissues. Based on the approach illustrated here, it should now be possible to introduce in mice a β4 mutation that abolishes signaling without affecting linkage to the cytoskeleton, and thus to examine if α6β4 signaling is necessary for homeostasis of adult mouse skin and intestine.
It has been speculated widely that the association of integrins with the cytoskeleton is necessary to stabilize adhesion to the ECM (Alberts et al., 1994). The phenotype of mice lacking the β4 tail provides a particularly vivid illustration of the importance of cytoskeletal anchorage for stable adhesion. Because of the association of β4 tail with the keratin cytoskeleton, stresses applied to the α6β4–laminin 5 bond may be distributed to the entire cytoskeleton. This view is consistent with the results of static and laminar flow adhesion assays and the observation that a significant fraction of tail‐less α6β4 appears to be pulled out of the membrane when the epidermis of mutant mice detaches from the basement membrane.
The phenotype of tail‐less β4 mice closely resembles that of humans affected by PA‐JEB, including the blistering of the skin and the detachment and degeneration of the pyloric epithelium. Therefore, the cell adhesion defect in PA‐JEB patients is most likely derived from the disruption of the transmembrane link between the basement membrane and keratin cytoskeleton mediated by α6β4. Our observation that α6β4 also regulates epidermal proliferation may explain another important feature of PA‐JEB. In contrast to individuals with classical JEB, which have a severe form of skin blistering caused by mutations in the α6β4 ligand laminin 5, but never display areas of skin hypoplasia or aplasia (Aberdam et al., 1994; Pulkkinen et al., 1994), newborns affected by PA‐JEB display extensive areas of skin aplasia, especially in the lower limbs (Vidal et al., 1995; Ruzzi et al., 1997). Since the production of differentiated keratinocytes appears to be normal in our mutant mice, it is possible that the precursor cell compartment in certain areas of human fetal skin is smaller than in mice so that the α6β4 mutation results in neonatal aplasia only in humans.
In conclusion, the results of this study indicate that the intracellular interactions mediated by the β4 cytoplasmic domain are necessary for stable adhesion of stratified epithelia to the basement membrane and for proper cell cycle control in both simple and stratified epithelia. These findings reveal that normal cells are also exquisitely anchorage‐dependent in vivo and help to understand the molecular basis of the human blistering skin disease PA‐JEB.
Materials and methods
Targeted deletion of the β4 cytoplasmic domain
A 7 kb fragment of the mouse gene β4 was isolated by screening a 129Sv library with a human cDNA probe complementary to the sequence encoding the transmembrane domain. Site‐directed mutagenesis was used to introduce a stop codon followed by a novel EcoRI site immediately after the sequence encoding the transmembrane domain. The EcoRI site was then used to subclone an SV40 polyadenylation signal followed by a neomycin resistance expression cassette downstream of the stop codon. Finally, the modified fragment was introduced into the targeting vector pPNT (Tybulewicz et al., 1991) in two steps (left arm 5 kb and right arm 1.8 kb). Prior to electroportation in ES cells (Swiatek and Gridley, 1993), the targeting vector was linearized by digestion with NotI. Positively transfected cells which had undergone homologous recombination were selected in 0.5 mg/ml G418 and 0.2 mM gancyclovir and identified by Southern blotting. Six distinct ES cell lines were found to carry the expected mutation, and four were injected into blastocyst‐stage C57BL/6 mouse embryos which were then transplanted into the uteri of pseudopregnant C57BL/6 mice. Two lines produced extensively chimeric male mice which were crossed to C57BL/6 females. Heterozygous offspring were then used to generate embryos homozygous for the targeted deletion. Embryos were genotyped by PCR using genomic DNA isolated from the forelimb. The primers used for amplification were: right arm primer, 5′‐GATCTTCCAGCGGACTGTGTC‐3′; neomycin cassette primer, 5′‐GCTCCCGATTCGCAGCGCCATCG‐3′; and wild‐type specific primer, 5′‐AGAGGATGGTGTCCTCCTCA‐3′. The wild‐type allele yielded a 0.5 kb fragment and the mutant allele gave rise to a 0.8 kb fragment.
Total RNA was extracted from mouse skin and hybridized to a 0.5 kb cDNA probe complementary to sequences in the β4 ectodomain. The probe was generated by amplifying reverse‐transcribed RNA from mouse PAM keratinocytes with the following primers: 5′‐CTGTGAGCAAGGAAGTT‐3′ and 5′‐CGTGTAGAGCGACTGCTGGTT‐3′.
The rabbit antiserum SE144 recognizes the γ2 subunit of laminin 5 (Vailly et al., 1994) and mAb 4E10 the β2 subunit of laminin 1, 2, 6 and 8 (Engvall et al., 1986). The rat anti‐mouse α6 mAb GoH3 and hamster anti‐mouse β1 mAb HMβ1‐1 were purchased from Pharmingen (Los Angeles, CA). The rabbit antiserum to a GST fusion protein comprising amino acids 31–217 of human β4 has been described (Mainiero et al., 1997). mAb 346‐11A reacts with the ectodomain of mouse β4 (Kennel et al., 1989). The antibodies to BPAG2 were raised by immunizing rabbits with a GST fusion protein comprising aminoacids 1–250 of mouse BPAG2 and affinity purified on the antigen. MAb5E reacts with BPAG1 (Shimizu et al., 1991) and mAb 121 with HD1/plectin (Hieda et al., 1992). Affinity‐purified rabbit antibodies to synthetic peptides from mouse filaggrin, loricrin, keratin 1, keratin 5 and involucrin were characterized previously (Calautti et al., 1995). The anti‐Ki‐67 mAb MM1 was produced at the Monoclonal Antibody Facility of Sloan Kettering Institute. Affinity‐purified antibodies to a GST fusion protein comprising full‐length p27Kip were described previously (Casaccia‐Bonnefil et al., 1997). Secondary antibodies were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA).
Skins from wild‐type and homozygous mutant E18.5 embryos were extracted in RIPA buffer containing protease inhibitors. Total epidermal proteins were normalized by immunoblotting with anti‐keratin 1 antibodies. Equivalent amounts of total epidermal proteins were immunoprecipitated with the anti‐α6 mAb GoH3 and subjected to immunoblotting with the antiserum to the β4 ectodomain (Mainiero et al., 1997).
Primary mouse keratinocytes were isolated as described (DiPersio et al., 1997). For static adhesion assay, primary keratinocytes from wild‐type and mutant E18.5 embryos were pre‐incubated with 10 μg/ml of anti‐β1 mAb HMβ1‐1, to block the adhesive function of α3β1, and plated for 30 min in the presence of the same antibody on microtiter wells coated with the indicated concentrations of purified laminin 5. For laminar flow adhesion assay (Brieher et al., 1996), primary keratinocytes from wild‐type and mutant E18.5 embryos were pre‐incubated with 10 μg/ml anti‐β1 mAb HMβ1‐1 and drawn into a glass capillary coated with 10 μg/ml purified laminin 5. The cells were allowed to bind under static conditions for 30 min in the presence of the same antibody and then subjected to increasing flow rates. The number of cells remaining adherent after each step was evaluated microscopically.
Immunohistochemistry and lactase assay
Skin and intestine samples were harvested from embryos and embedded in either OCT or paraffin. Tissue sections (10 μM) were stained with hematoxylin and eosin or subjected to either immunofluorescence or imunoperoxidase staining. Lactase activity was assayed on tissue sections as previously described (Stoward and Everson Pearse, 1991).
Note added in proof
A recent study provides further evidence that integrin‐mediated Shc signalling regulates cell proliferation in vivo [, in press].
We thank V.Soares and W.Mark of the Transgenic Mouse Facility, K.Manova of the Molecular Citology Facility, and N.Lampen of the Electron Microscopy Facility for their invaluable help; E.Calautti, G.P.Dotto and C.M.DiPersio for their advice on the isolation of primary keratinocytes; W.M.Brieher for assistance in the laminar flow assays; P.Rousselle and R.E.Burgeson for providing purified laminin 5; E.Colautti, G.P.Dotto, E.Engvall, T.Hashimoto, S.J.Kennel, A.Koff, G.Meneguzzi, K.Owaribe, M.Park and F.Watt for antibodies; and C.Blobel and B.Gumbiner for comments on the manuscript and discussions. This work was supported by DAMD grant 17‐94‐J4306 and NIH grants R01‐CA58976 and P30‐CA08748. C.M. was on leave of absence from the Istituto Nazionale della Nutrizione in Rome. F.G.G. is an Established Investigator of the American Heart Association.
- Copyright © 1998 European Molecular Biology Organization