Cloning and Complete Primary Structure of the Mouse Laminin α3 Chain DISTINCT EXPRESSION PATTERN OF THE LAMININ α3A AND α3B CHAIN ISOFORMS

We have isolated and characterized overlapping cDNA clones encoding the α3A and α3B chains of mouse laminin 5. Sequence analysis of the cDNA for the α3B predicts a polypeptide of 2541 amino acids (279,510 Da) comprising a truncated short arm and a carboxyl-terminal long arm common to the laminin α chains identified thus far. The short arm of the α3B chain harbors two alternating epidermal growth factor-like domains and two globular domains. The amino-terminal globular domain, thought to mediate interactions with molecules of the extracellular matrix, shows no significant homology to any globular domain at the tips of the known laminin isoforms. The α3A cDNA predicts a polypeptide of 1711 amino acids (186,230 Da) that substitutes a short sequence of 43 amino acids for the short arm seen in the α3B isoform and displays 77% conservative homology to the α3Ep chains of the adhesion ligand epiligrin. Northern and Western blot analyses of skin and lung epithelial cells demonstrated the tissue-specific expression of the laminin α3A and α3B isoforms, and in situ hybridization on mouse embryos revealed a focal localization of α3B in areas of the central nervous system.

We have isolated and characterized overlapping cDNA clones encoding the ␣3A and ␣3B chains of mouse laminin 5. Sequence analysis of the cDNA for the ␣3B predicts a polypeptide of 2541 amino acids (279,510 Da) comprising a truncated short arm and a carboxyl-terminal long arm common to the laminin ␣ chains identified thus far. The short arm of the ␣3B chain harbors two alternating epidermal growth factor-like domains and two globular domains. The amino-terminal globular domain, thought to mediate interactions with molecules of the extracellular matrix, shows no significant homology to any globular domain at the tips of the known laminin isoforms. The ␣3A cDNA predicts a polypeptide of 1711 amino acids (186,230 Da) that substitutes a short sequence of 43 amino acids for the short arm seen in the ␣3B isoform and displays 77% conservative homology to the ␣3Ep chains of the adhesion ligand epiligrin. Northern and Western blot analyses of skin and lung epithelial cells demonstrated the tissue-specific expression of the laminin ␣3A and ␣3B isoforms, and in situ hybridization on mouse embryos revealed a focal localization of ␣3B in areas of the central nervous system.
Laminins are noncollagenous components of basement membranes that mediate cell adhesion, growth, migration, and differentiation. These cross-shaped molecules constitute a family of proteins consisting of three individual polypeptide chains joined together in a long arm as coiled-coil amphipatic ␣-helices linked by interchain disulfide bonds. The amino-terminal domain of each of the three chains forms a distinct short arm (reviewed by Tryggvason, 1993).
Laminin chain variants with specific patterns of temporal and spatial expression have been identified in different species. All these isoforms are highly homologous, in that their short arms are comprised of globular domains and characteristic epidermal growth factor-like domains, and their long arms consist of sequences of heptad repeats (Timpl et al., 1979). On the basis of their primary structure deduced from sequence data and homology to the polypeptides that compose laminin 1, the laminin chains characterized thus far have been classified as ␣, ␤, or ␥ chains . Only chains belonging to a different class combine into a trimeric molecule presenting with a large globular domain G contributed by the carboxyl terminus of the ␣ chain.
Epithelial cells express specific laminin isoforms. Laminin 5 was initially identified by a monoclonal antibody that stains subsets of basement membranes (Verrando et al., 1987). The protein is associated with the anchoring filaments, thread-like structures connecting the hemidesmosomes to the lamina densa of the dermal-epidermal junction (Verrando et al., 1987;Rousselle et al., 1991). Laminin 5 is composed of three distinct chains of 165 kDa (␣3), 140 kDa (␤3), and 105 kDa (␥2). This mature species derives from a cell-associated molecule as a result of two extracellular processing events that generate the ␣3 and the ␥2 chains from distinct 200-and 155-kDa precursor polypeptides, respectively (Marinkovich et al., 1992a). The laminin ␣3 chain is immunologically related to a distinct laminin 190-kDa ␣ chain synthesized by keratinocytes that interacts with a ␤1 and a ␥1 chain to form laminin 6. Laminin 6 and laminin 5 appear to form a complex that functions as a cell adhesion ligand for integrins ␣6␤4 and ␣3␤1 (Carter et al., 1990;Marinkovich et al., 1992b). In amnios and fetal skin, the 190-kDa laminin ␣ chain associates also with a ␤2 and a ␥1 chain to form laminin 7 (Wewer et al., 1994).
Mutations in the genes encoding laminin 5, including its ␣3 chain, have been shown to underlie the junctional forms of epidermolysis bullosa, a recessive inherited skin disorder characterized by dysadhesion of the epidermis from dermis (Kivirikko et al. (1995) and references therein; Vidal et al. (1995)).
Screening for cDNA clones from a human keratinocyte expression cDNA library using a polyclonal antibody against the ␣3 chain of laminin 5 identified two species of mRNA transcripts. 1 Partial sequence analysis predicts two polypeptides identical to the ␣3 EpA and ␣3 EpB chain isoforms of the adhesive ligand epiligrin (Ryan et al., 1994), an anchoring filament component shown to mediate basal cell adhesion by interacting with integrin ␣3␤1 in focal adhesion and with integrin ␣6␤4 in hemidesmosome adhesion structures (Carter et al., 1991). The cDNA encoding the ␣ chain of epiligrin predicts two distinct polypeptides with identical COOH-terminal domains, homologous to domain IϩII and domain G of laminin ␣ chains, and totally divergent amino-terminal domains. The isoform ␣3 EpB substitutes a short arm, thus far uncharacterized, for the truncated amino-terminal domain of the ␣3 EpA counterpart (Ryan et al., 1994).
In this study, we report the complete cDNA sequences of the ␣3 chains of murine laminin 5, which demonstrate that the laminin ␣3B chain harbors a short arm with unique structural features. We also provide evidence that the laminin ␣3B and ␣3A isoforms display distinct expression patterns.

MATERIALS AND METHODS
Isolation and Analysis of cDNAs-A mouse lung cDNA library (Zap, Stratagene) was screened with a random-primed 32 P-labeled (prime-it, Stratagene) cDNA clone MN97 that encodes a region of the IϩII domain of the mouse laminin ␣3 chain . 12 positive phage clones were excised into pBluescript SK ϩ (Stratagene). The largest cDNAs, M100 (2.2 kb), 1 MR9 (1.5 kb), and M2C3 (3.3 kb), were completely sequenced on both strands using a Sequenase kit (Pharmacia Biotech Inc.). To obtain additional clones, the cDNA library was rescreened at low stringency using the human cDNA NA12, which codes for the 3Ј-end of the human laminin ␣3 chain, (Vidal et al., 1995) as a radioactive probe. A positive clone MZ6b (2.5 kb) containing the 3Ј-end of the complete nucleotide sequence of the cDNA was thus identified. cDNA clones representing extensions of cDNA M2C3 were isolated by PCR amplification of the mouse cDNA expression library. The primers used for 5Ј-extension were as follows: 5Ј-CAGTAGCAACACACTC-CTTA-3Ј (left) and 5Ј-CCAGGAGCACACTTGTC-3Ј (right), which correspond to a sequence in the vector T7 promoter and to a 5Ј-end sequence of cDNA MR9, respectively. The primers used for 3Ј-extension were as follows: 5Ј-CAGTAGCAACACACTCCTTA-3Ј (left) and 5Ј-TAGCCTGTGCCTTCAAAGTA-3Ј (right), which correspond to a 3Ј-end sequence of cDNA M2C3 and to a 5Ј-end sequence of cDNA MZ6b, respectively. After purification and subcloning of the amplification products into PCRTM-II vector (TA cloning kit, InVitrogen), clones M22 (960 nucleotides) and M0.2 (254 nucleotides) were completely sequenced. Clones representing the 5Ј-end of the complete nucleotide sequence of the laminin ␣3A and ␣3B cDNAs were obtained by 5Ј-rapid amplification of cDNA ends (5Ј-RACE kit; Life Technologies, Inc.). Briefly, 200 ng of poly(A ϩ )-selected RNA from the epithelial cell line PAM212 (Yuspa et al., 1980) was reverse transcripted using the primer 5Ј-TCGCAGTCATCACATTCTT-3Ј. The first strand DNA was amplified by PCR using a 5Ј-RACE kit as devised by the manufacturer with the provided sense primer (Anchor Primer) and the specific antisense primer 5Ј-CTGTGTTCCTGTGTATCCGG-3Ј, which corresponds to a sequence of cDNA M2C3. The 491-base pair PCR product was subcloned into the TA cloning PCRTM-II vector (InVitrogen) resulting in clone PR6, which contains the complete 5Ј-end of the laminin ␣3A isoform. Following a similar procedure, poly(A ϩ ) RNA from mouse lung was retrotranscribed into cDNA using primer 5Ј-GACAGAAGGAGGCAAG-GAAGGAACC-3Ј. PCR amplification was made using the Anchor Primer of the 5Ј-RACE kit and a specific antisense primer 5Ј-TTCA-CAATCACCTCAGT-3Ј, which corresponds to a 5Ј-end sequence of cDNA M22. The resulting MR10 clone (1145 base pairs) contains the complete 5Ј-sequence of the cDNA for the laminin ␣3B isoform. All of these clones were further sequenced on both strands. Primer extension analysis using the 5Ј-most fragment of clone MR10 and poly(A ϩ ) RNA prepared from mouse lung tissue showed that the 5Ј-untranslated sequence did not extend beyond the 5Ј-region of this cDNA.
Cell Culture-PAM212, a mouse transformed epidermal cell line, was cultured in Eagle's minimal essential medium supplemented with 10% fetal calf serum.
Northern Blot Analysis-Total RNA was isolated from adult mouse lung tissues and from cultures of actively growing PAM212 cells (Chomczynski and Sacchi, 1987). 20 g of poly(A) ϩ -enriched RNA was electrophoresed on 1.0% denaturing agarose gels, transferred onto nylon membranes (Hybond-N ϩ , Amersham), and hybridized with radioactive probes obtained by 32 P random priming of the cDNA fragment PR6H, which corresponds to the 230-base pair HindIII restriction product of cDNA PR6, the full-length cDNAs M100 and M22 inserts.
In Situ Hybridization Analysis-Sense and antisense probes specific to laminin transcripts ␣3A and ␣3B were obtained by labeling cDNAs PR6H, M22, and MZ6b with digoxigenin-uridine triphosphate (Boehringer Mannheim, France). In situ hybridization on mouse fetal tissues was performed using a method devised to detect mRNA transcripts in neural tissue (Schaeren-Wiemers and Gerfin-Moser, 1993).

Identification of cDNA Clones Encoding Mouse Laminin ␣3
Chain-Screening of a mouse lung cDNA library with a radioactive cDNA coding for the ␣3 chain of human laminin 5 identified cDNA clones corresponding to the full-length mouse lami- The scale is shown in nucleotides. B, structure of the ␣3A and ␣3B polypeptides with domain numbering in Roman numerals, according to Sasaki et al. (1988). The EGF-like repeats composing the cysteine-rich domains are represented by open boxes, numbered according to Sasaki et al. (1988). The COOH-terminal domain G is depicted by shaded boxes. nin ␣3 chains (see "Materials and Methods" and Fig. 1A). Sequence analysis revealed that these cDNA clones represent two distinct transcripts, designated as ␣3A and ␣3B, homologous to the epiligrin ␣3 EpA and ␣3 EpB chain isoforms, respectively (Ryan et al., 1994). The schematic structure of overlapping clones is depicted in Fig. 1A.
Nucleotide and Amino Acid Sequences of Laminin Isoform ␣3B-The complete amino acid sequence of the cDNA for the mouse laminin ␣3B chain is shown in Fig. 2A. The cDNA sequence contains an open reading frame (ORF) of 7704 nucleotides flanked by 5 nucleotides of 5Ј-untranslated sequence and 368 nucleotides of 3Ј-untranslated sequence. The 3Ј-non-coding sequence contains a polyadenylation signal (AAUAAA) located 16 nucleotides upstream of the poly(A) tail. The ORF encodes a protein of 2568 amino acids. The first in-frame ATG is in favorable context for initiation of translation (Kozak, 1991) and precedes a stretch of 26 hydrophobic amino acids typical of a signal peptide. According to the rule of Von Heijne (1986), a cleavage site was predicted following Gln-27. After cleavage of the signal peptide, the protein consists of 2541 residues with a predicted molecular mass of 279,510 Da. The mature protein has 11 putative N-linked glycosylation sites (Asn-X-Ser/Thr), and the molecular mass of the glycosylated peptide is estimated to exceed 300,000 Da.
Domain Structure of Laminin ␣3B Chain-The laminin ␣3B chain is comprised of a short arm of 1056 amino acids, a long arm comprising a rod-like region of 589 residues and a carboxyl-terminal globular domain G of 920 residues ( Fig. 2A). Recently, Ryan et al. (1994) reported the partial cDNA sequence for the ␣3 EpB of epiligrin, which resulted identical to the sequence of the ␣3B chain of human laminin 5. (Vidal et al., 1995). Since the cDNA sequence for the amino-terminal region of human laminin ␣3B chain is not available, we compared the domain structure of the short arms of mouse laminin chains ␣3B and ␣1. Alignment with the amino acid sequence of mouse laminin ␣1 chain reveals that the polypeptide chain ␣3B harbors a truncated amino-terminal end missing the most aminoterminal domains V and VI. Therefore, the short arm of the chain comprises the cysteine-rich EGF-like domains IIIa (residues 889-1057) and IIIb (residues 498 -699), which are predicted to have rigid rod-like structures, and domains IVa (residues 700 -888) and IV (residues 27-497), which are predicted to form globular structures (Fig. 1B).
No significant homology is found between regions of the large amino-terminal domain IV of the ␣3B chain and sequences in the amino-terminal domains of the laminin isoforms characterized thus far. In particular, the conserved sequences WWQS and Y(Y/F)YX 7 (G/R)G, located in the amino-terminal domain VI of most of the laminin chains (Sasaki et al., 1988;Hunter et FIG. 2. A, deduced amino acid sequence of mouse laminin ␣3B chain (upper line) aligned with the partial sequence of human ␣3 EpB (lower line). Double horizontal traits between the compared sequences underlined amino acid identities and single horizontal traits indicate conservative substitutions. The arrow shows the putative signal peptide cleavage site. Cysteine residues are boxed, and the potential N-linked glycosylation sites (NX(S/T)) are indicated by full circles. Amino acid sequences with a putative biological interest are underlined. Asterisk delimits the carboxyl-terminal sequences common to laminin ␣3A and ␣3B chains. Domains are boxed and labeled on the right. The sequence of the mouse ␣3B chain is available from EMBL under accession number X84014. B, nucleotide sequence of the cDNA encoding the aminoterminal domain specific to mouse laminin ␣3A chain (upper line) and deduced amino acid sequence (middle line). The amino acid sequence of the human epiligrin ␣3 EpA chain is also reported (lower line) (Ryan et al., 1994). Conserved amino acid residues are indicated by a horizontal line and differing residues by the appropriate one-letter code. Missing or mismatching amino acid residues are depicted by hatched squares. The putative cleavage site of the peptide signal is indicated by a triangle. The mouse ␣3A sequence is available from EMBL under accession number X84013. al., 1989;Beck et al., 1990;Kusche-Gullberg et al., 1992;Gerecke et al., 1994;Vuolteenaho et al., 1994;Wewer et al., 1994), are not found. Domain IV of laminin ␣3B chain has no significant homology with domain IV of laminin ␣1 and ␣2 chains. However, it displays 28% homology (42.6% if conservative changes are included) with domain IVЉ (residues 872-1374) of Drosophila laminin ␣ chain (Kusche-Gullberg et al., 1992). On the contrary, the globular domain IVa of the ␣3B chain (residues 700 -888) displays 19.8% homology to domain IVa of mouse laminin ␣1 (residues 1143-1344) and is 12 amino acids shorter. Moreover, the EGF-like domains IIIb of laminin ␣3B chain shows 47.2% homology with its counterpart in laminin ␣1 chain and is 249 amino acids shorter ( Table I). The best alignment is obtained with a sequence overlapping EGFs 7-11 of domain IIIb of laminin ␣1 chain (between positions 981 and 1142) (Sasaki et al., 1988). Domains IIIa of laminin chains ␣3B and ␣1 are 39.5% homologous. In the ␣3 chain, domain IIIa retains the four EGFs that constitute domain IIIa in the ␣1 chain (Sasaki et al., 1988). However, EGF 4 comprises only 6 cysteine residues, which are found in conserved positions (Sasaki et al., 1988). The size of the different domains of the short arm of mouse laminin ␣3B chain and their sequence homology with the corresponding domains of mouse and Drosophila ␣ chains are summarized in Table I.
Domain Structure of Laminin ␣3A Chain Isoform-The mouse laminin isoforms ␣3A and ␣3B present totally divergent 5Ј-ends and share identical amino acid sequences downstream of position 901 of the ␣3B cDNA ( Fig. 2A). The 5Ј-end region of laminin ␣3A chain is encoded by the cDNA clone PR6 and comprises 43 NH 2 -terminal amino acids specific to this chain that display 67.4% identity to the homologous 5Ј-amino acid sequence of the human ␣3A counterpart (Fig. 2B). The Gln residue at position 44 (Gln-901 in laminin ␣3B chain) is the first amino acid common to both ␣3A and ␣3B isoforms ( Fig.  2A). The full-length ␣3A cDNA (5563 nucleotides) comprises a 5Ј-untranslated region of 62 nucleotides and an ORF (5133 nucleotides), beginning with a Met codon surrounded by sequences fitting the eukaryotic translation start sites (Kozak, 1991). The initiation methionine precedes an appropriate signal sequence of 17 amino acids with consensus cleavage site following residue Glu-18 (Von Heijne, 1986) (Fig. 2B, arrow). The ORF encodes a protein of 1711 amino acids with nine consensus sites for N-linked glycosylation. The mature peptide has a predicted mass of 186,230 Da. The mass of the glycosylated peptide is estimated at 214,000 Da.
Differential Expression of Laminin Variants ␣3A and ␣3B-Since previous studies on the tissue distribution of the murine laminin ␣3 chain were performed using probes specific to the peptide COOH-terminal domains common to both ␣3A and ␣3B variants , we investigated whether cDNA probes for the distinct amino-terminal domains of the ␣3 chain isoforms detected expression patterns specific to each polypeptide. We first assessed the expression rate of the ␣3A and ␣3B transcripts by Northern blot analysis of mRNA purified from mouse lung and skin epithelial cells. Hybridization performed with the radioactive cDNA M100, which encodes for the rodlike domain common to both ␣3A and ␣3B chains, identified faint bands corresponding to transcripts with a size ranging between 5.5 kb in skin cells and 8.0 kb detected only in the lung (Fig. 3A). Using cDNA M22 as a radioactive probe specific for the ␣3B chain, a unique 8.0-kb band was detected in lung extracts (Fig. 3B), whereas using cDNA PR6, which encodes the   3. Expression of laminin ␣3 chain isoforms in lung and skin epithelial cells. 20 g of poly(A) ϩ RNA from adult lung tissue (lane 1) and epithelial cell line PAM212 (lane 2) were successively hybridized with 32 P-labeled cDNAs probes M100, which codes for a peptide common to both laminin ␣3A and ␣3B chains (panel A), M22, which is specific to the transcript ␣3B (panel B), and PR6H, which is specific to the ␣3A transcript (panel C).
amino-terminal domain of ␣3A, only the 5.5-kb band was specifically observed in epidermal cells (Fig. 3C). These results therefore indicated that some epithelia may express only one of the laminin ␣3 chain isoforms.
To verify this possibility, we further investigated the expression of laminin ␣3A and ␣3B isoforms at protein level. Western analysis was realized on total extracts prepared from mouse skin and lung using the polyclonal antibody SE152 specific to domain IϩII of the two mouse ␣3 chain isoforms . In skin extracts, the antibody detected a band with an apparent mass of 200 kDa and a 150 -165-kDa band doublet (Fig. 4), which is the electrophoretic migration pattern characteristic of the precursor and mature forms of laminin ␣3A chain (Marinkovich et al., 1992a;Aberdam et al., 1994). In lung extracts, antibody SE152 reacted with a single band with an apparent mass of 280 -300 kDa (Fig. 4), which is a value concordant with the estimated molecular weight of the polypeptide encoded by the full-length laminin ␣3B chain cDNA. It was thus clearly demonstrated that immunoreactivity to the polyclonal antibody SE152 in mouse skin and lung correlated with the presence in these tissues of mRNA for the laminin ␣3A and ␣3B chains, respectively. These results are therefore consistent with a cell type-specific expression of laminin ␣3A and ␣3B chain isoforms.
Focal Distribution of Laminin ␣3A and ␣3B Chain Isoforms-We then determined the tissue distribution of laminin ␣3A and ␣3B chains by in situ hybridization on mouse tissues using RNA probes specific for each isoform. According to our previous results , transcripts for both laminin ␣3A and ␣3B chains were detected in the basal membrane of the upper alimentary tract and urinary and nasal epithelia. The ␣3A chain appeared prominently expressed in the skin and, specifically, in hair follicles (Fig. 5A) and developing neurons of the trigeminal ganglion (13.5 days postcoitum) (Table II). Strong expression of the ␣3B chain was detected in the salivary glands and teeth, where the presence of ␣3A transcripts was also noticed. Conversely, ␣3B transcripts were exclusively found in the bronchi and alveoli, in the stomach and intestinal crypts, in the whisker pads (Fig. 5D), and in the central nervous system (Table II). In the brain, strong hybridization signals were seen in the telencephalic neuroectoderm (Fig. 5F) and a transient (13.5 days postcoitum), and focalized expression was also observed in the thalamus, the Rathke's pouch, and the periventricular subependymal germinal layer (Fig. 5F).

DISCUSSION
In the present study, we relate the cloning of cDNAs coding for the full-length ␣3A and ␣3B isoforms of mouse laminin ␣3 chain, and we demonstrate the tissue-specific distribution of these laminin variants.
Sequence data reveal that the ␣3B chain isoform (300 kDa) substitutes the amino-terminal short arm with two alternating cysteine-rich domains and two globular domains for the short amino-terminal peptide found in the ␣3A counterpart (200 kDa). These observations are concordant with previous results reporting that human laminin ␣3A chain harbors a short arm, consisting of a reduced thread-like structure comprised of four EGF-like repeats, and a long arm, identical to the COOHterminal regions of the ␣3B chain isoform (Ryan et al., 1994). Apart from a restrained region matching 29.7% of the amino acid sequence of domains III and IVa of laminin ␣1 chain, the short arm of mouse laminin ␣3B chain presents no homology with laminin ␣1 and ␣2 chains. The most amino-terminal domain of the polypeptide displays a weak sequence similarity with domain IVЉ of Drosophila laminin ␣ chain, which has been suggested to arise from the fusion of a duplicated domain IVЈ of laminin ␤1 chain (Kusche-Gullberg et al., 1992). However, in this, the two laminin ␣ chains differ because no homology is found between the ␣3B chain and laminin ␤ isoforms.
The electron microscopy images of laminin 5 purified from keratinocytes depict the molecule as a rod-like entity missing the short arms characteristic of classical laminins or the globular structures of the laminin ␣3B chain short arm. It has thus been suggested that the ␣3B transcript corresponds to the ␣ chain polypeptide of laminin 6 (K-laminin) (Ryan et al., 1994). However, lines of evidence suggest that the ␣ chains of these two laminins are distinct isoforms. First, the ␣ chain of laminin 6 is a truncated polypeptide lacking the amino-terminal short arm (Marinkovich et al., 1992b). Second, the deduced molecular mass (300 kDa) of polypeptide ␣3B is inconsistent with the estimated mass of the ␣ chain of laminin 6 (190 kDa) (Marinkovich et al., 1992b). No evidence for processing of the ␣ chain of laminin 6 that could account for this discrepancy has thus far been provided. Third, synthesis of laminin 6 in H-JEB patients is not affected by mutations in the LAMA3 gene, resulting in hampered expression of ␣3A and ␣3B transcripts (Vidal et al., 1995). Therefore, detection of the 300-kDa polypeptide corresponding to the ␣3B transcript in lung epithelia, where the laminin ␣3A chain is not detected, supports the assumption of the existence, and the coexistence in some epithelia, of laminin 5 isoforms harboring distinct ␣3 chains.
Thus far, information on laminin 5 has been gathered from studies performed on the protein secreted by epidermal keratinocytes. In the lamina lucida of the dermal-epidermal junction, laminin 5 immunolocalizes with the anchoring filaments of hemidesmosomes (Rousselle et al., 1991;Verrando et al., 1993) and codistributes with integrin ␣6␤4, the transmembrane receptor associated with hemidesmosomes (Sonnenberg et al., 1991). The major importance of laminin 5 for the formation of the hemidesmosomal adhesion complex and for the cohesion of the dermal-epidermal junction is deduced from the observation that in H-JEB an impaired expression of the protein correlates with abnormality in hemidesmosome structures and extensive skin blistering (Verrando et al., 1991). Herlitz JEB is also characterized by disadhesion of gastrointestinal and lung epithelia in which hemidesmosomal complexes have not been described (Jones et al., 1994). Therefore, the possibility exists that, in gut and basal membranes, laminin 5 incorporates the ␣3B chain to associate with morphological structures distinct from hemidesmosomes via the globular domains of the short arm of the ␣3B chain. This would be consistent with the detection of laminin ␣3B chain in brain tissues where laminin 5 may assume roles other than epithelial adhesion.
The intense expression of the ␣3B transcript in neuroectoderm and cerebellum confirms previous studies on expression of laminin 5 during organogenesis, suggesting a role in brain and nerve development . Laminin ␣1 chain harbors the peptide SIKVAV that mediates cell attachment, migration, and neurite outgrowth (Tashiro et al., 1989). Since in the mouse laminin ␣3 chain the peptide is not conserved, the sequence divergence within this area may reflect a functional difference in the laminin ␣3 chain. The murine laminin ␣3A chain, however, is focally expressed in the developing trigeminal nerve, which strengthens the idea that this polypeptide may play a role in the migration and polarization of motor neurons. The laminin ␤2 chain, which directs the growing axons of intraspinal commissural neurons via the motor neuron-selective adhesive site (LRE), is expressed in the central nervous system (Sanes et al., 1990;Aberdam et al., 1994). Interestingly, the human and murine ␣3 chains contain LRE sites, which suggests that this isolaminin may be physiologically active in the development of areas of the central nervous system. However, the possible function of laminin 5, or that of laminin isoforms comprising ␣3 chains, in the development of the central nervous system deserves further investigations and accurate clinical evaluation of JEB patients with mutations in the LAMA3 gene.
From our results, it seems likely that structural variants of the ␣3 chain may contribute to regulate diverse functions of laminin 5. Cloning of the cDNAs for the murine laminin ␣3 chains sets the stage for experiments on gene disruption in mice embryonic stem cells for the analysis of the specific role of the laminin ␣3A and ␣3B chain isoforms.

TABLE II
In situ hybridization of mouse tissue sections ϩϩ, very strong; ϩ, strong; Ϯ, weak; and Ϫ, no staining. For this survey, in situ hybridization was performed on sections of tissue obtained from 13.5-and 17.5-day embryos (estimated gestation ages) with PR6H, M22, and MZ6b RNA probes corresponding to regions specific to ␣3A, ␣3B, and common to both, respectively.