Reduced expression of the epithelial adhesion ligand laminin 5 in the skin causes intradermal tissue separation.

Laminin 5, the major keratinocyte adhesion ligand, is found in the lamina lucida subregion of the epidermal basement membrane of the skin, where it colocalizes with the anchoring filaments. Mutations in the genes encoding laminin 5 cause junctional epidermolysis bullosa, an inherited skin blistering disease characterized by abnormal hemidesmosomes and cleavage of the lamina lucida leading to epidermal detachment. In this work we describe the genetic basis of a new subtype of lethal inherited epidermolysis bullosa associated with reduced skin reactivity to laminin 5, presence of mature hemidesmosomes, and intradermal cleavage of the skin. The epidermolysis bullosa patients were heterozygous for a nonsense mutation (Q896X) and a splice site mutation (764-10T-->G) in the gene (LAMC2) for the gamma2 chain of laminin 5. The nonsense mutation causes accelerated decay of the corresponding mRNA, while the splice site mutation results in maturation of a cryptic wild-type gamma2 mRNA leading to reduced expression of wild-type laminin 5. In vitro studies using the probands' keratinocytes showed that secretion of reduced amounts of functional laminin 5 in the patient, although permitting formation of hemidesmosomes, fail to restore efficient cell adhesion. Our results provide the first evidence that laminin 5 contributes to the firm adhesion of the epithelial basement membrane to the underlying stroma. They also show that a low expression level of laminin 5 induces assembly of mature hemidesmosomes in vivo but fails to assure a stable cohesion of the dermal-epidermal junction.

Laminin 5, the major keratinocyte adhesion ligand, is found in the lamina lucida subregion of the epidermal basement membrane of the skin, where it colocalizes with the anchoring filaments. Mutations in the genes encoding laminin 5 cause junctional epidermolysis bullosa, an inherited skin blistering disease characterized by abnormal hemidesmosomes and cleavage of the lamina lucida leading to epidermal detachment. In this work we describe the genetic basis of a new subtype of lethal inherited epidermolysis bullosa associated with reduced skin reactivity to laminin 5, presence of mature hemidesmosomes, and intradermal cleavage of the skin. The epidermolysis bullosa patients were heterozygous for a nonsense mutation (Q896X) and a splice site mutation (764؊10T3 G) in the gene (LAMC2) for the ␥2 chain of laminin 5. The nonsense mutation causes accelerated decay of the corresponding mRNA, while the splice site mutation results in maturation of a cryptic wild-type ␥2 mRNA leading to reduced expression of wild-type laminin 5. In vitro studies using the probands' keratinocytes showed that secretion of reduced amounts of functional laminin 5 in the patient, although permitting formation of hemidesmosomes, fail to restore efficient cell adhesion. Our results provide the first evidence that laminin 5 contributes to the firm adhesion of the epithelial basement membrane to the underlying stroma. They also show that a low expression level of laminin 5 induces assembly of mature hemidesmosomes in vivo but fails to assure a stable cohesion of the dermal-epidermal junction.
Epidermolysis bullosa (EB) 1 is a heterogeneous group of inherited diseases characterized by epithelial fragility to mechanical friction resulting in blisters and erosions of the integument. Extracutaneous involvement includes erosions of squamous and transitional epithelia, pitted enamel, alopecia, and nail dystrophia. The different forms of EB are associated with molecular defects in components of the dermal-epidermal junction (1). The dermal-epidermal junction is a complex basement membrane (BM) ultrastructurally composed of four compartments: the ventral plasma membrane of the basal keratinocytes and the associated hemidesmosomes; the lamina lucida, which includes the anchoring filaments; the lamina densa; and the upper papillary dermis containing the anchoring fibrils (2). About 20 biochemical components have so far been identified that localize in the different compartments of the dermal-epidermal junction. Among them type IV collagen, the backbone of all the basement membranes, is the major component of the lamina densa; type VII collagen is the prominent component of the anchoring fibrils; and laminins 5, 6, and 10 localize both in the lamina lucida and the lamina densa (3). The hemidesmosome-anchoring filament complex and the anchoring fibrils form a specialized attachment structure common to the stratified and simple epithelia exposed to the external environment (3).
Based on ultrastructural data and antigen mapping, EB has been divided into three major clinical subtypes according to the level of the blister plane (4). In EB simplex, tissue separation occurs within the basal keratinocytes as a result of genetic mutations in keratins 5 and 14 or in the hemidesmosomal component plectin. In junctional EB (JEB), blisters arise within the lamina lucida, with the basal epidermal cells at the roof of the blister and the lamina densa at its base. Tissue cleavage results from mutations in the extracellular adhesion ligand laminin 5 and the two transmembrane components of hemidesmosomes, integrin ␣ 6 ␤ 4 and collagen type XVII. In dystrophic EB, the split localizes within the papillary dermis below the lamina densa and results from mutations in collagen type VII. Immunostaining of EB skin using an adequate panel of antibodies may detect the abnormal expression of a specific BM component and identify the candidate gene(s) in the disease (4).
Laminin 5 (the laminin ␣3␤3␥2 heterotrimer) is specifically associated with the hemidesmosome-anchoring filament complexes (5,6). It has been reported to influence morphogenetic events (7,8), to affect cell motility (9,10), to enhance invasiveness in human cancers (11)(12)(13)(14)(15), and to control cell growth (16). In the skin, laminin 5 is synthesized and secreted into the extracellular matrix by the basal keratinocytes and constitutes a specific substrate for static attachment and for adhesion of proliferating and migrating cells. Null mutations in any of the three genes (LAMA3, LAMB3, and LAMC2) encoding laminin 5 cause the extensive epithelial disadhesion distinctive of lethal JEB (H-JEB, for Herlitz JEB) (17). Mutations leading to syn-thesis of an aberrant but partially functional laminin 5 result in mild JEB (nH-JEB, for non-Herlitz JEB), which is characterized by a limited tendency to skin blistering. In H-JEB, hemidesmosomes are absent or hypoplastic and/or reduced in number. In nH-JEB, hemidesmosomes may be reduced in number, lack the hemidesmosomal subbasal dense plate, and have rudimentary cytoplasmic plaques. Transfer of a curative transgene that rescues expression of functional laminin 5 in H-JEB keratinocytes restores hemidesmosome assembly and cell adhesion (18).
It has been demonstrated that laminin 5 mediates epithelial cell adhesion by interacting with integrins ␣ 6 ␤ 4 and ␣ 3 ␤ 1 through its COOH-terminal G-domain (19 -21). In blistered nH-JEB skin, immunoreactive laminin 5 distributes along the floor of the blister, which indicates interaction of the protein with components of the lamina densa and the papillary dermis. Furthermore, laminin 5 binds the NC-1 domain of collagen type VII (22,23) and associates with laminin 6 in the skin and laminin 7 in the amnion (24,25). Additional evidence supporting a structural function of laminin 5 in basement membrane stability includes immunochemical data, which suggests that initiation of skin basement membrane formation involves assembly of laminin 5 into laminin complexes and interaction with integrins ␤ 1 and ␤ 4 (26). However, the active role of laminin 5 in structuring of the basement membrane has not been substantiated by direct in vivo observations and remains hypothetical.
In this study, we have disclosed the molecular basis of an unusual form of lethal epidermolysis bullosa associated with novel genetic mutations in the LAMC2 gene. This form is characterized by a reduced production of wild type laminin 5, the presence of mature hemidesmosomes, and tissue separation underneath the lamina densa of the dermal-epidermal junction. Our results provide the first evidence that laminin 5 contributes to the firm adhesion of the epithelial basement membrane to the underlying stroma.

MATERIALS AND METHODS
Epidermolysis Bullosa Patients-The two EB probands were generated by healthy parents in a family with a history free of skin diseases and genetic disorders. Proband 1 developed extensive cutaneous blisters and erosions of the upper digestive tract shortly after birth. EB was diagnosed on the basis of ultrastructural analysis of involved skin that revealed severing of the dermal-epidermal junction. The infant died at the age of 6 months. Proband 2 was the result of a second pregnancy. At the 21st week of gestation, prenatal diagnosis performed on fetal skin biopsies by electron microscope examination detected epithelial detachment. The pregnancy was terminated. Pathologic examination of the fetus was performed immediately after delivery and confirmed the fragility of the skin and mucosa.
Biological Samples and Cells-Primary keratinocytes isolated from skin biopsies were cultured on lethally irradiated feeders of mouse 3T3-J2 fibroblasts in the presence of a mixture (3:1) of DMEM and Ham's F-12 medium (Life Technologies, Cergy-Pontoise, France). The medium contained 10% fetal calf serum, 5 g/ml insulin, 0.4 g/ml hydrocortisone, 0.1 nM cholera toxin, 10 ng/ml epidermal growth factor, 2 nM triiodothyronine, and 0.18 mM adenine (27). Total RNA was purified from cultured keratinocytes using the RNable™ extraction kit (Eurobio, Les Ullis, France). Genomic DNA was extracted from peripheral blood following standard techniques.
Electron and Immunofluorescence Microscopy-Ultrastructural examination of noninvolved skin and mechanically induced blisters was performed as described (28). Ultrathin sections were examined with a JEOL EM 1200 transmission electron microscope. Indirect immunofluorescence microscopy was performed on 5-m sections of frozen skin samples and on cultures of primary keratinocytes (28). The samples were examined using a Zeiss Axiophot epifluorescence microscope.
Western Blot Analysis-For Western blot analysis of the extracellular matrix (ECM) deposited by the proband's keratinocytes, subconfluent cell cultures were detached by treatment with 10 mM EDTA in phosphate-buffered saline (PBS). The ECM was then scraped from the culture support in the presence of 2% SDS, 10% glycerol, 100 mM ␤-mercaptoethanol, 50 mM Tris-HCl, pH 6.8. The samples were briefly sonicated and cleared by centrifugation at 10,000 ϫ g for 5 min at 4°C. The protein concentration was determined by the Bradford method (Bio-Rad). 25 l of the ECM and serial dilutions of the original solution in PBS were electrophoresed in a 6% acrylamide SDS-polyacrylamide gel and blotted on nylon filters (35). The antibody-antigen complex was revealed using the ECL Western blotting kit (Amersham Pharmacia Biotech). Quantitation of autoradiograms was performed by densitometric scanning using the Bioprofil software program (Vilber Lourmat, France).
Detection and Verification of Mutations-To search for mutations in LAMC2 individual exons were PCR-amplified using 100 ng of genomic DNA as a template and oligonucleotide primer pairs synthesized on the basis of the flanking intronic sequences (37). Specifically, for amplification of the 230-bp DNA fragment comprising the 189-bp exon 7 (nucleotides 764 -953 in the cDNA; GenBank TM number U31184) and containing the maternal mutation 764Ϫ10T3 G, the following primers were used: 5Ј-TCATAGCCCAAAGTGCATA-3Ј (forward) and 5Ј-AG-GCTTGTTGAACAAATGAA-3Ј (reverse) (37). For amplification of the 152-bp exon 18 (nucleotides 2601-2753 in the cDNA; GenBank TM number U31195) containing the paternal mutation Q896X, the primers were 5Ј-GGTGTGGTAACTGGTAAGCA-3Ј (forward) and 5Ј-TATTCATATG-GTGGGCATTC-3Ј (reverse). The PCR conditions were as follows: 95°C for 5 min followed by 30 cycles at 94°C for 30 s, 58°C for 45 s, and 72°C for 30 s. 5 l of the PCR products were used for heteroduplex analysis by conformation-sensitive gel electrophoresis (38) followed by direct automated sequencing of the PCR products containing a heteroduplex band (ABI, Foster City, California). The presence of mutations Q896X and 764Ϫ10T3 G in members of the family and unrelated healthy controls was assessed by direct sequence analysis. Occurrence of mutation 764Ϫ10T3 G was further investigated in 100 alleles of unrelated healthy controls by allele-specific oligonucleotide hybridization as described (36). The allele-specific oligonucleotides were 5Ј-GCTGTGAAA-CAGAGTTTTA-3Ј and 5Ј-GCTGTGAAACCGAGTTTTA-3Ј, corresponding to the wild type and mutant LAMC2 sequences, respectively.
RT-PCR-Total RNA isolated from the control and the patient's cultured keratinocytes was reverse transcribed in a volume of 25 l using avian myeloblastosis virus reverse transcriptase as recommended by the manufacturer (Promega, Madison, WI). One l of each reaction product was used to amplify overlapping regions of the cDNAs spanning the open reading frame of the laminin ␥2 chain. Specifically, for amplification of the 439-bp cDNA fragment (nucleotides 588 -1027; Gen-Bank TM number X73902) (31) including the 23-bp deletion generated by the maternal mutation 764Ϫ10T3 G, the oligonucleotide primers were 5Ј-GCAGCTTCTGCAGAATACAGT-3Ј (forward) and 5Ј-AGATTCCG-CAGTAACCTTCG-3Ј (reverse). The PCR conditions were as follows: 94°C for 3 min followed by 30 cycles at 94°C for 20 s, 58°C for 45 s, and 72°C for 45 s. The amplification products were submitted to direct automated sequencing.
To isolate the wild-type RNA transcripts of the maternal LAMC2 allele, 1 l of the reverse transcription product was submitted to allelespecific long range PCR amplification using the Expand PCR system (Roche Molecular Biochemicals) to generate a 1.9-kilobase cDNA fragment (nucleotides 763-2705). The sense primer (5Ј-CAAATTCTTGG-GAATCAACAG-3Ј) was homologous to the sequence deleted in the maternal ␥2 cDNA bearing the mutation 764Ϫ10T3 G. The antisense primer (5Ј-CCAGTTTCCCAGATTCTTCTG-3Ј, nt 2685-2705) was homologous to the maternal ␥2 cDNA fragment harboring guanine 2688 (which in the paternal ␥2 cDNA is substituted by a polymorphic adenosine). The PCR conditions were as follows: 94°C for 3 min, 94°C for 20 s, 52°C for 45 s, 68°C for 1 min 30 s (10 cycles), 94°C for 20 s, 52°C for 45 s, and 68°C for 2 min (25 cycles with an increment of 20 s in extension time at each stage). Direct sequence analysis of the amplimer was performed using the primer 5Ј-GATTCAGTGTCTCCGCTTCA-3Ј (nucleotides 2557-2576).
Quantification of Laminin 5 Deposited in the ECM by Keratinocyte Cultures-Laminin 5 deposition on plastic cell culture substrate was quantified by enzyme-linked immunosorbent assay. Cells (2 ϫ 10 3 , 5 ϫ 10 3 , and 1 ϫ 10 4 /well) were seeded in multiwell plates (96 wells; TPP, Zurich, Switzerland) and incubated for 48 h at 37°C in a humidified atmosphere in the presence of 5% CO 2 . The wells were then washed twice in PBS and detached as previously described by Delwel et al. (39). Briefly, cells were incubated overnight at 4°C in Hanks' buffered saline solution containing 20 mM Hepes, 20 mM EDTA, 1 mM EGTA, and a general protease inhibitor mixture (ICN Biomedicals, Aurora, OH) and then dislodged by pipetting. The matrices were washed with PBS, incubated for 10 min with PBS plus 1% Triton X-100 to remove cell debris, and saturated for 1 h at 37°C in PBS plus 2% bovine serum albumin (BSA). Each well was incubated for 2 h at room temperature with 50 l of PBS plus 2% BSA containing mAb GB3 (10 g/ml). The plates were washed three times with PBS and incubated 1 h at room temperature with 50 l of PBS plus 2% BSA containing 1 g/ml of anti-mouse HRP mAb (Dako) per well. After three washes with PBS, 50 l of a solution of o-phenylenediamine dihydrochloride (Sigma) were added to each well for 15 min in the dark as devised by the supplier. The reaction was stopped by adding a 2 M solution of H 2 SO 4 , and color yields were determined at 490 nm in an enzyme-linked immunosorbent assay reader (Dynatech, Guyancourt, France).
Cell Detachment and Adhesion Assays-Keratinocytes (3 ϫ 10 4 cells/ well) were plated in the presence of serum-free medium in 96-well tissue culture plates (Polylabo, Strasbourg, France). The wells had previously been coated with 0.5% bovine serum albumin fraction V (Sigma) or 10 g/ml Engelbreth Holm-Swarm Mouse Sarcoma matrigel (Sigma) by overnight adsorption at 4°C, washed twice with PBS, and saturated for 1 h at room temperature with 0.5% heat-inactivated BSA in PBS. Cells were allowed to adhere for 1, 16, or 24 h at 37°C. After washing with PBS, the attached cells were fixed with 3% paraformaldehyde for 20 min at room temperature, washed three times with PBS, and stained with 0.5% crystal violet in 20% methanol for 15 min. Excess dye was washed off with water, and the stained cells were eluted with 100 l of a solution containing 50% ethanol and 0.1 M sodium citrate, pH 4.2. Absorbance at 560 nm was read in a microplate manager (Bio-Rad). Nonspecific adhesion was evaluated by seeding the keratinocytes in wells coated with 0.5% BSA. Each assay point was derived in triplicate. For cell detachment assays, suspensions of secondary keratinocytes (6 ϫ 10 4 cells/cm 2 ) were seeded in tissue culture flasks and incubated for 48 h at 37°C. The cell cultures were then rinsed and treated with a solution of 0.05% trypsin, 0.01% EDTA (0.5 M), pH 8. The number of cells detached at different times was determined by collecting the supernatants of each cell culture and direct cell counting.
All measurements were performed in triplicate, and the results obtained from the three different experiments were expressed as the mean of S.D.

RESULTS
The Intradermal Cleavage of the Dermal-Epidermal Junction Is Associated with a Reduced Immunoreactivity of Laminin 5-The probands presented with the characteristic clinical phenotype of severe EB. Ultrastructural examination of nonlesional skin detected mature hemidesmosomes and well formed anchoring filaments connected to the lamina densa of the basement membrane zone (Fig. 1A). In fresh lesions of blistered skin, the cleavage plane localized below the lamina densa, and hemidesmosomes were well formed (Fig. 1B). Careful examination of the involved skin also detected lesions with the cleft lying within the lamina lucida and presenting normally shaped hemidesmosomes at the roof of the blister (Fig. 1C). Immunoreactivity of noninvolved skin to mAb GB3 against laminin 5 was strongly reduced compared with the control skin of nonaffected donors (Fig. 2, A and B). In contrast, immunoreactivity of integrin ␣ 6 (Fig. 2, G and H) and that of all of the major basement membrane components (not shown) was comparable with that detected in control skin. This finding suggested the possibility of a potential association of the disease with an altered expression of laminin 5. Indeed, labeling with the antilaminin ␥2 pAb SE144 was faint (Fig. 2, C and D), while staining with mAb K140 (Fig. 2, E and F) and pAb SE85 (data not shown) specific to the laminin ␤3 and ␣3 chains, respectively, was slightly decreased. These observations suggested that LAMC2 was the candidate gene in the condition. In the lesional skin, immunostaining of laminin 5, type IV collagen, laminin 1, and type VII collagen was located at the roof of the blisters (Fig. 3, A-D), which confirmed the sublamina densa severing of the basement membrane. In a few blisters, the cleavage plane partially localized within the lamina lucida, above the lamina densa (Fig. 3, insets).
The EB Probands Are Compound Heterozygous for Mutations in the LAMC2 Gene-The finding that the abnormal expression of laminin 5 is associated with the presence of well formed hemidesmosomes and a cleavage plane of the skin lying below the lamina densa is intriguing, because so far the altered expression of laminin 5 has always been associated with tissue separation within the lamina lucida and dysplastic hemidesmosomes (40,41). To confirm the direct implication of laminin 5 in the condition, the expression level of the laminin ␤3 and ␥2 chains was determined by Northern blot analysis of total RNA purified from keratinocytes isolated from the EB fetus. Compared with control keratinocytes, expression of the laminin ␥2 chain was 80% reduced in the patient's cells, whereas the steady state level of the laminin ␤3 transcripts was not affected (Fig. 4).
To identify possible pathogenic genetic mutations in LAMC2, the probands' genomic DNA was PCR-amplified using primers homologous to the intronic DNA sequences flanking each exon of LAMC2 (37). Conformation-sensitive gel electrophoresis revealed heteroduplex formation with the PCR products corresponding to exon 18 (not shown). Direct nucleotide sequencing of the amplimers detected a heterozygous C3 T transition at position 2686 of the laminin ␥2 cDNA (GenBank TM accession number X73902), which substitutes a termination codon TAA for glutamine 896 (CAG) (Fig. 5A). The mutation was designated Q896X. A conservative G3 A transition was also detected at the wobble position of the codon for glutamine 896 (Fig. 5A). Direct nucleotide sequencing of the PCR products amplified from the genomic DNA of the proband's parents showed that mutation Q896X and polymorphism 2688 G3 A co-segregate with the paternal allele (not shown).
Heteroduplex formation was also observed with the PCR products encompassing exon 7. Direct nucleotide sequencing of the PCR products detected an intronic T3 G transversion at position Ϫ10 of exon 7 (764Ϫ10T3 G) in the patient (Fig. 5B), in the affected fetus, and in the mother. To determine the effect of mutation 764Ϫ10T3 G at the mRNA level, the laminin ␥2 cDNA was amplified by RT-PCR using total RNA purified from the proband's keratinocytes. Direct nucleotide sequencing of the 439-bp cDNA fragment encompassing exons 6 and 7 of LAMC2 identified a wild type mRNA and an internally shortened transcript bearing a 23-nt out-of-frame deletion at the 5Ј-end of exon 7 (nt 764 -786 of the laminin ␥2 cDNA). This deletion results in a shift of the open reading frame and makes a downstream premature termination codon (TGA at nt 813-816; Fig. 5C). The internally deleted mRNA is generated by an abnormal splicing involving a putative cryptic acceptor splice site within exon 7 (cag/G, nt 784 -787) (Fig. 5D). The possibility that mutation 764Ϫ10T3 G constitutes a polymorphic variation in the population was excluded by allele-specific oligonucleotide screening of a panel of 100 unrelated control alleles (data not shown). Screening for mutations detected no additional base change either in the remaining LAMC2 cDNA sequences or introns 5-7. We therefore concluded that mutations Q896X and 764Ϫ10T3 G are causative for EB phenotype in the kindred. This conclusion was confirmed by the results of an early DNA-based prenatal testing performed in a subsequent pregnancy. Genotyping of the fetal genomic DNA prepared from a chorionic villous biopsy obtained at 10 weeks of preg-FIG. 2. Immunohistochemical analysis of the noninvolved skin from proband 1. In cryostat skin sections, staining of the basement membrane using mAb GB3 specific to native laminin 5 was attenuated and discontinuous (A) compared with healthy controls (B). Immunoreactivity of each individual chain of laminin 5 is also reduced with a strong attenuation of the fluorescent signal with the anti-␥2 pAb SE144 (C) and a less marked reduction with the anti-␤3 mAb K140 (E) and anti-␣3 pAb SE85 (not shown). In contrast, immunoreactivity to mAb GoH3 specific to the integrin ␣ 6 subunit is comparable in the patient (G) and in the control (H). Scale bar, 48 m.

FIG. 3. Immunohistochemical analysis of the lesional skin in proband 2.
Immunostaining to laminin 5 (A) collagen type IV (B), laminin 1 (C), and collagen type VII (D) is localized at the roof of the blisters, indicating that the cleavage plane of the basement membrane zone lies beneath the lamina densa. In a subungula blister, the basement membrane components localized at the floor of the lesion, indicating that the sewerage localizes within the lamina lucida (insets). Scale bars, 48 m. nancy showed that the fetus had inherited the wild type LAMC2 alleles. At delivery, the newborn was clinically unaffected.
The EB Keratinocytes Express Wild-type Laminin 5-Similar to the probands' epidermis, the probands' keratinocytes expanded in culture were weakly reactive to pAb SE144 (Fig. 6A) and to mAb GB3 (data not shown). Accordingly, the ECM deposited by these cells on the culture vessel was faintly stained compared with the ECM layered down by wild-type keratinocytes, indicating deposit of trimeric laminin 5. Expression of the ␥2 polypeptide by the proband's keratinocytes was assessed and quantified by immunoblotting of the ECM deposited by these cells on the Petri dishes using pAb SE144. Two hybridization bands were detected with the apparent electrophoretic mobility of the unprocessed (155 kDa) and extracellularly processed (105 kDa) ␥2 polypeptides (Fig. 6B) (25). By densitometric analysis, the intensity of the two bands was estimated to be 100-fold weaker than that of the corresponding bands detected in cultures of wild type keratinocytes.
In light of these results, we hypothesized that the maternal mutation 764Ϫ10T3 G permits synthesis of full-length ␥2 polypeptides that incorporate into laminin 5 heterotrimers. To verify this possibility, incorporation of native laminin 5 in the extracellular matrix layered down by the proband's keratinocytes was determined by enzyme-linked immunosorbent assay using mAb GB3. As shown in Fig. 6C, wild type keratinocytes efficiently deposited laminin 5 on the plastic culture substrate, while the proband's keratinocytes secreted barely detectable amounts of the protein. Taking advantage of the polymorphic nucleotide 2688 detected in the paternal LAMC2 allele, synthesis of a wild-type ␥2 mRNA was then verified by long range allele-specific RT-PCR amplification of the ␥2 RNA transcripts using total RNA purified from the probands' keratinocytes as a template. The antisense primer was specific to the maternal ␥2 cDNA fragment spanning nt 2686 -2705, and the sense primer was complementary to the 23-bp cDNA sequence absent in the internally deleted maternal ␥2 mRNA. Direct sequencing showed that the PCR amplification products (nt 763-2705 of the ␥2 cDNA) contained the guanine residue encoded at nt 2688 by the maternal LAMC2 allele and harbored the 23-bp sequence previously found to be deleted in the maternal transcripts. These results demonstrated that the maternal muta- FIG. 4. Northern analysis of the patient's keratinocytes. Northern blot analysis was performed using total RNA purified from primary cultures of keratinocytes obtained from the proband (P) and an unrelated control (C). The RNA was electrophoresed on denaturing agarose gels, blotted to a nitrocellulose sheet, and successively probed with 32 P-labeled cDNAs for the laminin ␤3 and laminin ␥2 chains. A reduction in the steady-state level of the ␥2 messenger RNA was noted in the proband's keratinocytes. The mutation was designated Q896X. In the mutated allele, a polymorphic G3 A transition was also detected at the wobble position (nt 2688) of the codon for glutamine 896 (asterisk). The presence of mutation Q896X in the EB kindred was assessed by nucleotide sequencing of the genomic DNA obtained from the patients' relatives. The mutated allele was detected in the probands' father and in the affected fetus. WT, wild type. B, maternal mutation. Direct nucleotide sequencing of the proband 230-bp PCR amplimer, spanning exon 7 of LAMC2 and its intronic boundaries, detected a heterozygous T3 G transversion at position Ϫ10 of intron 6 as shown by the sequence of the reverse strand (lower panel). The mutation was designated 764Ϫ10T3 G. Direct sequencing of the genomic DNA from the mother detected mutation 764Ϫ10T3 G in one allele. C, direct nucleotide sequencing of the cDNA reverse-transcribed from the laminin ␥2 of the patient identified a heterozygous out of frame 23-bp deletion (nucleotides 764 -787 of the ␥2 cDNA) as shown by the sequence of the reverse strand. D, schematic representation of the aberrant splicing of the ␥2 pre-mRNAs transcribed from the maternal LAMC2 allele. Mutation 764Ϫ10T3 G activates a cryptic splice site (AG, nucleotides 785 and 786 in the ␥2 cDNA sequence) within exon 7 and induces an out-of-frame internal deletion of 23 nucleotides. tion 764Ϫ10T3 G allows rescue of a minor form of wild type ␥2 pre-mRNA (Fig. 6D).
The Proband's Keratinocytes Display Reduced Adhesion Capacity in Vitro-Secretion of laminin 5 accounts for the assembly of mature hemidesmosomes in the patients' skin and promotes cell adhesion both in vivo and in vitro (42). To evaluate the adhesive capacity of the probands, keratinocyte cell suspensions were seeded either on EHS matrigel substrate or on plastic coated with BSA. Secondary ␥2-null keratinocytes isolated from an unrelated H-JEB (43) and wild type keratinocytes were used as controls. Similar to the normal keratinocytes, the number of the proband's cells sticking to the EHS matrigel substrate increased with time and at a comparable extent, while the percentage of adhering ␥2-null keratinocytes increased much slower. In the absence of matrigel, only the wild type keratinocytes efficiently attached to the culture substrate, indicating that the extracellular matrix layered down by the EB keratinocytes was inefficient in promoting cell adhesion on unprimed plastic culture substrate (Fig. 7A).
Because in the absence of the ECM culture substrate the adhesion of the proband's keratinocyte was comparable with that of the ␥2-null keratinocytes, the adhesive capacity of these cells was further assessed using a cell detachment kinetic assay (22). Exponentially growing cell cultures were treated with a trypsin/EDTA solution, and the number of cells dislodged at increasing intervals of time was evaluated by direct counting. The results showed that the proband's keratinocytes rapidly detached from the cell culture support. However, these keratinocytes displayed a higher adhesion capacity than the ␥2-null counterparts because only 12% of them were dislodged after 6 min and 47% after 8 min, while for the ␥2-null cells, values were 52 and 83%, respectively (Fig. 7B). At 15 min, however, 100% of the EB keratinocytes were detached compared with 25% of the wild type cells.
Taken together, these observations underscore the dramatic adhesion deficiency of the proband's keratinocytes but also indicate that production of scant amounts of laminin 5 plays a perceptible effect on cell adhesion that could explain the sublamina densa cleavage detected in the two EB patients.

DISCUSSION
In this study, we have examined two cases of a rare variant of inherited lethal EB. Blistering was associated with a reduced expression of laminin 5, as illustrated by the barely detectable immunoreactivity of the probands' skin to the monoclonal antibody GB3, which recognizes native laminin 5, and to the polyclonal antibody SE144 directed against the laminin ␥2 chain. Previously, lethal EB caused by hampered expression of laminin 5 has been associated with the assembly of sparse and/or incomplete hemidesmosomes, resulting in a pronounced tendency to dermal-epidermal tissue separation in the plane of the lamina lucida of the basement membrane zone. In these two patients, transmission electron microscopy detected normally formed hemidesmosomes, and blister formation predom- inantly localized beneath the lamina densa. Tissue separation at the basement membrane/dermis interface was confirmed by immunohistochemical staining of the lesional skin that mapped all of the major components of the basement membrane to the roof of the blisters. It is unlikely that artifacts caused the atypical cleavage of the basement membrane, because the ultrastructural and immunohistochemical analyses were performed using fresh mechanically induced blisters, and skin biopsy specimens were processed following well established procedures routinely used to study EB epidermis. The presence of structurally normal hemidesmosomes in a patient suffering from severe junctional EB has been previously reported in only one case (41). The patient died a few days after birth, and neither immunomapping of the dermal-epidermal junction nor genetic studies were performed. Therefore, here we present the first evidence that in patients with lethal EB, laminin 5 mutations may permit assembly of structurally normal hemidesmosomes and cause blisters with a sublamina densa cleft of the dermal-epidermal junction.
The two EB patients were found to be compound heterozygotes for two novel mutations in the LAMC2 gene. As attested by Northern blot analysis of total RNA purified from one of the probands' keratinocytes, the genetic defects led to reduced levels of ␥2 mRNA expression. One of the mutations was a nonsense mutation (Q896X) within the ␣-helical coiled-coil domain of the polypeptide, which predicted a truncation of the laminin ␥2 chain by 75%. It has been demonstrated that premature termination codons cause accelerated decay of the abnormal RNA transcripts and absent expression of the corresponding polypeptide (44). The shortened ␥2 polypeptide is not expected to incorporate into laminin 5, because the COOH-terminal coiled-coil domain I of the long arm that is crucial for assembly of the laminin heterotrimers is missing (Refs. 45 and 46 and references therein). In addition, Western blot analysis with keratinocyte protein extracts using the polyclonal antibody SE144 (31) was negative for the presence of the shortened ␥2 polypeptide with an expected size of ϳ100 kDa. It is therefore likely that the mRNA transcript containing the nonsense mutation undergoes rapid decay, which prevents synthesis of the aberrant polypeptide.
The other mutation was a single base pair substitution (764Ϫ10T3 G) at the 3Ј splice site of intron 6 that activates a putative cryptic splice site (CAGG) within exon 7 on the maternal LAMC2 allele. The aberrant splicing generates a mutated mRNA with an internal out-of-frame deletion that results in a downstream premature termination codon predicting truncation within the NH 2 terminus (domain IV) of the ␥2 chain short arm. Absent synthesis of the aberrant polypeptide due to mRNA decay is also likely.
Intronic mutations adjacent to the 5Ј splice site reduce the efficiency of mRNA splicing and activate cryptic splice sites (47). Legitimate splicing may, however, take place at a reduced rate, provided that spliceosome assembly is not abolished (48). In our patients, allele-specific RT-PCR amplification of the mRNA demonstrated that a wild-type ␥2 transcript is synthesized from the maternal allele containing mutation 764Ϫ10T3 G. According to these results and the faint immunoreactivity of skin and cultured keratinocytes to monoclonal antibody GB3, Western blot analysis of the extracellular matrix deposited by the proband's keratinocytes in a culture vessel detected laminin 5 molecules with a full-length ␥2 polypeptide. Rescue of wild type transcripts from pre-mRNA with intronic mutations has been documented previously (47). In particular, moderation of the skin blistering in two unrelated EB patients has been associated with intronic mutations in the genes COL7A1 and ITGB4, which allow expression of normal colla-gen type VII and integrin ␤ 4 polypeptides, respectively (36,49). In our patients, expression of cryptic wild type ␥2 mRNA restored secretion of functional laminin 5 molecules that mediated assembly of ultrastructurally normal hemidesmosomes but did not support epithelial adhesion.
Laminin 5 adheres efficiently to surfaces and primes the plastic and glass culture vessels for cell adhesion (50 -52). Lines of evidence have indicated that accumulation of the extracellular matrix ligand at a threshold level is required for efficient spreading and cell adhesion (42,53). The amount of ␥2 chain detected in the ECM produced in vitro by the proband's keratinocytes was 100-fold lower than in normal keratinocytes, and the ␥2 polypeptides were efficiently processed as judged from the respective intensity of the 155-and the 105-kDa bands visualized by Western blot analysis. The highly restricted production of ␥2 chain results in secretion of laminin 5-containing extracellular matrix unable to support stable adhesion. Indeed, adhesion assays showed that the probands' keratinocytes adhere poorly to plastic coated with BSA, while they attach efficiently to plastic precoated with EHS matrigel. This observation underscores the inability of scant amounts of laminin 5 to form a high affinity adhesion substrate in vitro and also provides a possible explanation for the unusual ultrastructural features of the patients' skin. Indeed, it has been shown that in healthy skin 90% of laminin 5 concentrates immediately under the hemidesmosomes, while the remaining 10% distributes along the basement membrane, where it is thought to promote stable epithelial-stromal attachment by interacting with laminin 6/7 (5,24,54). In our patients, scant expression of laminin 5 in the basement membrane apparently may favor its accumulation on top of the anchoring fibrils, which mediates efficient nucleation of hemidesmosomes, and limit its deposition in the interhemidesmosomal space and in the anchoring plaques of the papillary dermis (23). The sub-lamina densa cleavage of the blistered skin may thus reflect both the firm fastening of the anchoring fibrils to the lamina densa and the fragility of the deepest layers of the dermal-epidermal junction. On the other hand, the junctional cleavage noted in a number of blisters may result from insufficient deposition of laminin 5 underneath the hemidesmosome, which weakens the connection between the anchoring filaments and the lamina densa, as observed in JEB patients with mutations causing loss or synthesis of abnormal laminin 5.
In conclusion, our observations expand the spectrum of genetic heterogeneity in EB and also demonstrate the critical contribution of laminin 5 to the cohesion of the epithelial basement membrane. Furthermore, they underscore the need for a correct evaluation of the effect associated with the transfer of curative genes in EB. Phenotypic reversion of keratinocytes defective for proteins of the hemidesmosome-anchoring fibril complex has recently led to proposals for gene therapy of inherited EB (55). In this context and according to observations made on transgenic mice (56), here we show that assembly of ultrastructurally normal hemidesmosomes may not reflect a full recovery of cell adhesion but only partial rescue due to insufficient expression of the curative polypeptide.