The assembly of laminin-5 subunits.

Laminin-5 is a heterotrimer composed of α3, β3, and γ2 chains, produced by keratinocytes and the human squamous cell carcinoma line (SCC-25), and is one of the candidate proteins for the genetic lesion in junctional epidermolysis bullosa. Two-dimensional SDS-polyacrylamide gel electrophoresis (first dimension, nonreducing conditions; second dimension, reducing conditions) revealed that the immunoprecipitated laminin-5 from a SCC-25 cell fraction consisted of α3, β3, and γ2 monomers, a β3γ2 heterodimer, and an α3β3γ2 heterotrimer. The presence of the β3γ2 heterodimer, but not heterodimers containing an α3 chain and any of the other chains, was suggestive of assembly of laminin-5 proceeding from a β3γ2 heterodimer to an α3β3γ2 heterotrimer. We showed, by cotransfection experiments using full-length recombinant β3 and γ2 chains in a human cell line devoid of endogenous laminin-5, that stable heterodimers can be formed in the absence of α3 chain expression. In the SCC-25 cell fraction, the α3 monomer pool was the smallest of the monomers. Pulse-chase experiments using the cell fraction also indicated that the heterotrimer was assembled after a 10-min pulse and was nearly absent after a 24-h chase. These results are consistent with the synthesis of α3 being limiting for heterotrimer assembly, with rapid association of the α3 chain with β3γ2 heterodimers to form complete heterotrimers. Treatment with tunicamycin reduced the size of each of the laminin-5 subunits, indicating that all chains are glycosylated, but that N-linked glycosylation is not necessary for chain assembly and secretion.

Laminin-5 (kalinin/nicein) is an epithelium-specific laminin subtype and is a component of the anchoring filament of the lamina lucida in the basement membrane of the skin (1). The anchoring filament is the bridge between hemidesmosomes and the lamina densa region of the basement membrane and is believed to mediate the adhesion of the epithelium to the basement membrane. Laminin-5 is therefore one of the primary adhesion proteins holding together the epidermis and the dermis. Laminin-5 is initially synthesized in a cell-associated form, estimated to be 460 kDa, and is composed of three polypeptides: the 200-kDa (␣3), 145-kDa (␤3), and 155-kDa (␥2) chains (2). The three chains are presumed to form a cruciform structure where the chains are bound by disulfide linkages (2). The two heterotrimeric forms in keratinocyte cell-conditioned culture medium are derived from the cellular form by extracellular processing. A 440-kDa medium form results from the processing of the 200-kDa ␣3 chain to 165 kDa, while the 400-kDa form is derived from the 440-kDa form by extracellular processing of the 155-kDa ␥2 chain to 105 kDa (2).
Two models of laminin-1 chain assembly, possibly relevant to the steps in laminin-5 assembly, have been proposed. Peters et al. (3) and Morita et al. (4,5) observed a disulfide-linked ␤1␥1 heterodimer as a presumed intermediate and therefore suggested that the ␣1 chain was added at a later stage. Alternatively, Wu et al. (7) reported that initially laminin chains are assembled randomly.
Defects in laminin-5 result in defective anchoring filaments, causing blistering of the skin, and are now known to cause Herlitz junctional epidermolysis bullosa (HJEB), 1 an autosomal recessive disorder characterized by generalized blister formation at the level of the lamina lucida within the epidermal basement membrane (8,9). Recently, mutations in the ␥2 chain (10,11) and ␤3 chain (12) genes of laminin-5 have been reported in HJEB patients. Since laminin-5 is a heterotrimer, knowledge of the steps in the assembly of the complete protein may be important in future attempts to correlate specific chain mutations with laminin-5 dysfunction and ultimately with clinical phenotype. As a first step in understanding the pathophysiology of HJEB, we have characterized the subunit assembly of laminin-5.
Our assembly study was conducted in a squamous cell carcinoma line . This cell line produces laminin-5 that is indistinguishable from that produced by normal human keratinocytes (2). We determined the steps in laminin-5 chain assembly using endogenous laminin-5 and checked our conclusions using exogenous chains expressed from full-length cDNAs. We conducted immunoprecipitation reactions with general and chain-specific antibodies and used two-dimensional gels to resolve subunit association.

EXPERIMENTAL PROCEDURES
Antisera-Fusion proteins derived from cDNAs encoding the central domains of the ␥2 (13) and ␤3 (14) proteins was used to raise anti-␥2 and anti-␤3 antisera in rabbits. The cDNA clones were subcloned into the pGEX prokaryotic expression vector (Pharmacia Biotech Inc.). The fusion protein was prepared upon induction, purified according to manufacturer's instructions, and used as an antigen for immunization of rabbits. Polyclonal anti-laminin-5 antiserum was kindly provided by * This work was supported by NIAMS Grants AR41045-03 and AR19537. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Partially supported by a scholarship from the Uehara Memorial Foundation.
Biosynthetic Labeling Study and Immunoprecipitation-Squamous cell carcinoma line SCC-25 (ATCC CRL1628) was cultured in 50% Ham's F-12 medium, 50% DMEM supplemented with 0.5 g/ml hydrocortisone, and 10% fetal bovine serum. Cells grown to subconfluency in 60-mm plastic dishes were incubated in methionine-and cystine-deficient DMEM for 2 h and then cultured in deficient DMEM with 100 Ci/ml protein labeling mixture ([ 35 S]methionine and [ 35 S]cystine) (Du-Pont NEN) for 2 h (cell) or 24 h (medium). After the first 6 h of the total 24-h incubation used for the preparation of the medium fraction, 0.1 volume of fresh DMEM was added to the deficient DMEM to prevent amino acid exhaustion. Cell and medium fractions were processed for immunoprecipitation, which was performed as described previously (15). Briefly, cell layers were harvested with a cell scraper and ice-cold radioimmune precipitation buffer (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 250 M phenylmethylsulfonyl fluoride, 1 mM Nethylmaleimide, 2 mM L-methionine, 2 mM L-cystine, 0.3% Nonidet P-40, 0.05% Triton X-100, 0.3% sodium deoxycholate, 0.1% bovine serum albumin) containing 0.1% SDS and 0.1 M iodoacetamide. 250 mM phenylmethylsulfonyl fluoride and 0.1 M iodoacetamide were added to the culture medium and centrifuged at 2000 rpm to remove cells and debris. Medium and cell lysate were cleaned by a mixture of preimmunized rabbit serum, anti-rabbit IgG-conjugated agarose beads (Sigma), and gelatin-Sepharose (Pharmacia Biotech Inc.) for 1 h. Immunologically reactive laminin-5 was precipitated from precleared medium and cell fractions by a 16-h incubation with a mixture of anti-rabbit IgG-conjugated agarose beads and specific antibodies. After incubation, precipitates were pelleted by centrifugation at 2500 rpm for 10 min and washed with radioimmune precipitation buffer (medium sample) or radioimmune precipitation buffer containing 0.1% SDS (cells). After five washes, the pellets were mixed with SDS sample buffer, heated to 95°C for 3 min, and analyzed by SDS-PAGE.
Pulse-Chase Kinetic Analysis of Laminin-5 Biosynthesis-Dissociated SCC-25 cells were seeded at 10 6 cells/60-mm dish and allowed to attach in complete growth medium for 24 h and then incubated in methionine-and cystine-deficient DMEM for 2 h. Cells were pulsed with 100 Ci/ml protein labeling mixture in deficient DMEM for 10 min and then chased by incubation in complete growth medium. In these experiments, cells and conditioned medium were removed after 0, 0.5, 1.5, 3, 6, and 24 h of chase in complete medium and then processed for immunoprecipitation. The precipitates of immunoreactive laminin-5 forms were fractionated by SDS-PAGE.
Interaction between Recombinant ␤3 and ␥2 Subunits in 293 Cells-We constructed full-length ␤3 and ␥2 chain cDNAs that encode the entire open reading frame from previously published partial cDNA clones (13,14) using overlapping internal restriction endonuclease sites. The constructs were first placed in a Bluescript vector (Stratagene) and then inserted into the eukaryotic expression vector pCPS (DNA Factory, 415-FASTDNA). The expression vector contains the transcriptional control regions of human cytomegalovirus, including the cytomegalovirus enhancer and promoter, and contains an intron preceding the multiple cloning site. The inserted cDNA is followed by the polyadenylation site and the transcription terminator of the late region of SV40. The plasmids were designated pCPS␤3 and pCPS␥2. These plasmids were transfected into the human embryonic kidney cell line 293 (ATCC CRL1573) by calcium phosphate-mediated transfection (16). Briefly, 24 h before transfection, 293 cells were replated at a density of 5.0 ϫ 10 6 cells/60-mm plate in DMEM with 10% fetal bovine serum. Calcium phosphate precipitates containing 5 g of each plasmid were prepared. When one plasmid or a no-plasmid-added control was used, pCPS was added to adjust the total DNA concentration to 10 g/0.5 ml of medium. The cells were incubated with the precipitate for 4 h at 37°C, followed by a glycerol shock treatment for 2 min. The medium was then replaced by fresh medium, and the cells were grown for an additional 48 h. Transfected 293 cells were preincubated in methionineand cystine-free DMEM for 2 h and labeled in deficient DMEM with 100 Ci/ml protein labeling mixture for 2 h. Cell fractions were divided and processed for immunoprecipitation with anti-␤3 or anti-␥2 antibody.
Effect of Tunicamycin on Laminin-5 Biosynthesis-SCC-25 cultures seeded at the same density were pretreated in deficient medium for 2 h with no additives, followed by 3 g/ml tunicamycin (Sigma). Cells were labeled by replacing the medium with a methionine-and cystine-free medium containing 100 Ci/ml protein labeling mixture with or without tunicamycin. After labeling the cell fraction for 2 h and the medium for 24 h, cell and medium fractions were processed for immunoprecipitation.

Biosynthesis and Assembly of the Laminin-5 Protein in
Using nonreducing conditions to analyze the conditioned medium (Fig. 1c, lanes L, ␤, and ␥), two secreted forms of the laminin-5 molecule, the 440-kDa (␣Ј␤␥) and 400-kDa (␣Ј␤␥Ј) forms, were observed (␣Ј and ␥Ј are the processed forms of the ␣ and ␥ chains, respectively, whereas the ␤ chain is not processed). Monomers or dimers are probably not secreted since no bands are seen even after longer exposures of the autoradiographs corresponding to the molecular masses of either monomers or dimers. Under reducing conditions, these same samples containing both forms of the heterotrimer resolved into four bands, corresponding to 165-kDa processed ␣3 (␣Ј), 145-kDa ␤3 (␤), 155-kDa ␥2 (␥), and 105-kDa processed ␥2 (␥Ј) (Fig.  1d, lanes L, ␤, and ␥). All three antibodies also precipitated an additional band of 250 kDa under reducing conditions (Fig. 1d,   Fig. 2a (lower right panel). The signal intensity of the monomers under nonreducing conditions was ␣ Ͻ ␥ Ͻ ␤. A weaker signal for ␣3 was reproducibly observed as compared with ␤3 (␤) or ␥2 (␥) monomers, perhaps indicating that the intracellular pool of uncombined ␣3 is smaller than that of the other chains. We cannot rule out the possibility of preferential precipitation of the ␤ or ␥ chain by our general antisera against laminin-5.

Kinetics of Laminin-5 Biosynthesis, Subunit Assembly, and
Secretion-SCC-25 cells were pulse-labeled for 10 min with [ 35 S]methionine and [ 35 S]cystine and then chased with nonradioactive culture medium for intervals ranging from 0 to 24 h (Fig. 3a). In the cell fraction, after a 6-h chase, monomers nearly disappeared, but ␤␥ dimers and ␣␤␥ heterotrimers persisted. After a 24-h chase, most of the immunoreactive laminin-5 was in the heterotrimeric form without even small amounts of the heterodimer remaining. In the cell samples under reducing conditions (Fig. 3b), the ␣3 subunit signal seemed to decay faster than the other two subunits as shown by the 6-and 24-h chase samples. Note that there was no indication of any processing events leading to altered mobilities on SDS-PAGE regardless of the length of chase.
In the medium, after a 1.5-h chase, a 440-kDa band was observed under nonreducing conditions (Fig. 3c). The smaller 400-kDa form of laminin-5 appeared after a 3-h chase, and its intensity steadily increased up to the 24-h chase period. A band corresponding to the 460-kDa cellular form was not detected clearly. The same samples under reducing conditions contained three bands corresponding to the subunits of the 440-kDa form when observed after a 3-h chase (Fig. 3d). The processed ␥2 subunit belonging to the 400-kDa form appeared after a 6-h chase. Two additional protein bands of approximately 180 and 200 kDa appeared during the chase period (Fig. 3d). These bands also appeared in the 3-and 6-h chase lanes and decreased by 24 h (Figs. 1d and 2b) (reducing conditions). The two bands are most likely intermediates of the processed ␣3 chain since a 200-kDa protein is immunoblotted by an anti-laminin-5 antibody and is present in pulse-chase experiments of keratinocyte-conditioned medium (4). A small amount of laminin-5 containing the partially processed ␣3 chain may be unresolvable from the 440-and 400-kDa laminin-5 bands on nonreduced SDS-PAGE (Fig. 2b). The 250-kDa fibronectin band was Formation of a ␤␥ Heterodimer Using Recombinant ␤3 and ␥2 Chains-We used full-length cDNA expression vector constructs for ␤3 and ␥2 to determine if the ␤␥ dimer is readily assembled as an intermediate in the formation of laminin-5. The target cell for these experiments was the human embryonic kidney cell line 293, which does not express laminin-5 as determined by immunoblotting of the cell fraction with antilaminin-5 antibody (data not shown; Fig. 4a, lane 13). The assembly of recombinant ␤3 (r␤) and ␥2 (r␥) chains was conducted in the complete absence of the ␣3 chain.
The recombinant ␤3 chain was immunoprecipitated from ␤3 cDNA-transfected 293 cells using anti-␤3 antibody (Fig. 4a,  lane 3), but not from untransfected cells or from cells transfected with the vector alone (Fig. 4a, lanes 1 and 2). As expected, anti-␤3 antibody did not recognize r␥ in ␥2 cDNA-transfected 293 cells (Fig. 4a, lane 4); however, under reducing conditions, immunoprecipitates from ␤3and ␥2-co-transfected 293 cells using the same antibody showed both r␤ and r␥ (Fig. 4a, lane 5), providing evidence for the formation of a ␤␥ heterodimer. The recombinant chains showed the same migration on SDS-PAGE under reducing conditions as endogenous ␤3 and ␥2 from SCC-25 cells (Fig. 4a, lane 6). Similar results were obtained when the same set of transfection experiments were immunoprecipitated with anti-␥2 antibody. As expected, ␥2 antiserum did not react with r␤, only with r␥ (Fig.  4a, lanes 9 and 10), and coprecipitated r␤ and r␥ from cotransfected 293 cells (lane 11), providing further evidence of ␤␥ heterodimer formation. Our anti-laminin-5 antibody was able to recognize both individual recombinant chains (Fig. 4a, lanes  15 and 16). In this case, recognition of the two recombinant chains together in the cotransfection is not necessarily the consequence of coprecipitation (Fig. 4a, lane 17).
Under nonreducing conditions, similar results were obtained, although multimers of r␤ and r␥ were evident in the cell extracts from ␤3or ␥2-transfected 293 cells (Fig. 4b, lanes 3, 5,  10, 11, 16, and 17). These high molecular mass bands dissociated into bands of the same size as r␤ or r␥ monomers upon electrophoresis under reducing conditions (Fig. 4a). Significantly, a heterodimer of r␤ and r␥ was detected that could be distinguished from the multimer bands in the extracts from ␤3and ␥2-cotransfected 293 cells (compare lanes 3 and 5 and lanes  10 and 11).
To further clarify the identity of the presumed ␤␥ heterodimer, we performed two-dimensional SDS-PAGE analysis with the samples from cotransfected cells. The heterodimers of r␤ and r␥ have a similar molecular mass (300 kDa) to the homodimers of either r␤ (290 kDa) or r␥ (310 kDa). We resolved the presumed heterodimer on a second dimension under reducing conditions into its component r␤ and r␥ chains. Anti-␤3 antibody reacted with the r␤ monomer (Fig. 5a, black asterisk) and heterodimer (black and white arrows). The high molecular mass multimers dissociated into r␤ (Fig. 5a, black arrowheads) and r␥ (white arrowheads).
Anti-␥2 antibody recognized both the r␥ monomer, which resolved into two bands (Fig. 5b, white asterisks), and heterodimer (black and white arrows). The r␥ monomer resolved into two bands in the first dimension under nonreducing conditions, probably due to the presence of an intramolecular disulfide bond in the r␥ monomer. In the second dimension (reducing conditions), the high molecular mass multimers dissociated into only r␥ (Fig. 5b, white arrowheads), with no r␤containing species visible. Although the two antibodies showed different reactivity to the multimers, cotransfected 293 cells definitely contain monomers (r␤ and r␥), multimers, and the heterodimer (r␤r␥).
Glycosylation of Laminin-5, Subunit Assembly, and Secretion-Tunicamycin inhibits the synthesis of dolichol phosphate N-acetylglucosamine (an intermediate required for the synthesis of N-glycosylated glycoprotein) and has been used to study the role of oligosaccharide side chains in protein assembly (18). For the purposes of the discussion below, we labeled the tunicamycin-treated unglycosylated subunit precursors with a "p." In the presence of tunicamycin, monomers (␣p, ␤p, and ␥p), dimers (␤p␥p), and heterotrimers (␣p␤p␥p) migrated as lower molecular mass species as compared with the untreated normal cell lysate samples under nonreducing conditions (Fig. 6a, compare lanes N and T in the top panel). These presumably correspond to unglycosylated precursors as a result of tunicamycin treatment. A similar observation was made under reducing conditions, where each individual chain migrated as a lower molecular mass species as compared with the untreated normal sample (Fig. 6a, compare lanes N and T in the lower right  panel). The assembly of unglycosylated laminin-5 subunits was assessed by two-dimensional SDS-PAGE. The heterotrimer (␣p␤p␥p), under nonreducing conditions, dissociated upon reduction into a mixture of the unglycosylated subunits (␣p, ␤p, and ␥p), as did the heterodimer (␤p␥p). These results suggest that in the absence of glycosylation, the assembly of laminin-5 subunits occurs in the same way as in untreated normal cells.
Finally, we assessed the role of glycosylation in the secretion and subsequent processing of laminin-5 by conducting a similar analysis as described above using the medium of tunicamycintreated cells. In SDS-PAGE of the medium from the tunicamycin-treated cells under nonreducing conditions (Fig. 6b, lane T  in the top panel), two bands migrating as lower molecular mass species as compared with the untreated normal medium frac-tions were observed, corresponding to heterotrimers (␣pЈ␤p␥p and ␣pЈ␤p␥pЈ). Even after overexposure of this gel, no dimer and monomer bands were seen. Under reducing conditions, the same samples showed four fast migrating bands of unglycosylated proteins (Fig. 6b, ␣pЈ, ␤p, ␥p, and ␥pЈ in the lower right panel). On two-dimensional SDS-PAGE analysis, the larger band of first dimension electrophoresis dissociated into three proteins, processed ␣3 (␣pЈ), unprocessed ␤3 (␤p), and unprocessed ␥2 (␥p) precursors. The lower signal dissociated into processed ␣3 (␣pЈ), unprocessed ␤3 (␤p), processed ␥2 (␥pЈ) precursors. Both the tunicamycin-treated cell and medium fractions showed a small band estimated to be 80 kDa under reducing conditions (Fig. 6, a and b, lane T in the lower right panels). Since both anti-␤3 and anti-␥2 reacted with this band on immunoprecipitation (data not shown), it may be a product of degradation of the heterotrimer, due to an increased instability of unglycosylated precursors as compared with the normally glycosylated chains. DISCUSSION We have characterized the subunit assembly of a novel tissue-specific laminin variant, laminin-5, formerly known as nicein or kalinin (1,2). Laminin-5 is present at epithelial-stromal interfaces and is the primary component of the anchoring filaments associated with hemidesmosomes and therefore is crucial to the attachment of basal epithelial cells to the basement membrane zone. Defects in laminin-5 have been shown to be a cause of a lethal skin disease, Herlitz junctional epidermolysis bullosa (10 -12). We utilized the squamous cell carcinoma line SCC-25 as a source of constitutively high levels of laminin-5 rather than cultured primary keratinocytes because of the tendency of the keratinocytes to differentiate, resulting in a down-regulation of laminin-5. As is the case for a variety of multimeric proteins, we observed a specific order of addition for the assembly of individual chains into heterotrimeric laminin-5. In the cell fraction containing laminin-5 prior to secretion, we observed monomers of each of the three subunits (␣3, ␤3, and ␥2), a ␤3␥2 heterodimer, and an ␣3␤3␥2 heterotrimer. The presence of a ␤␥ heterodimer, but the absence of any other heterodimeric species, was suggestive of an ordered assembly of laminin-5 proceeding from the ␤␥ heterodimer to the heterotrimer by the addition of the ␣3 chain.
We employed a eukaryotic expression vector to further test the possibility of a stable association of the ␤ and ␥ chains into a heterodimer as an intermediate step in the assembly of the laminin-5 heterotrimer. Earlier use of recombinant proteins to study assembly of other laminin isotypes has been restricted to the use of the E8 fragment of mouse Engelbreth-Holm-Swarm laminin (19), prokaryotic expressed Engelbreth-Holm-Swarm laminin (20), and synthetic peptides (21) in vitro. Expression of recombinant proteins has been an important tool in the study of subunit assembly of a variety of proteins, however, including the interaction of fibrinogen subunits (22,23). Here we utilized complete cDNA expression vectors for the ␤ and ␥ chains to establish that the two chains stably heterodimerize in the absence of the ␣3 chain. Our experiment also rules out the possibility that the ␤␥ heterodimer is a breakdown product of the heterotrimer since assembly occurs in the absence of the ␣3 chain. We used 293 cells for our transfection experiments because they contain no endogenous laminin-5 chains and are transfectable at high efficiencies and yield high levels of protein expression when exogenously added genes are driven by the cytomegalovirus promoter. The fact that the human embryonic kidney cell line 293 will express the ␤ and ␥ chains and allow their interaction makes the involvement of any tissuespecific factors in ␤␥ heterodimer formation unlikely.
The ␣3 monomer pool was observed to be the smallest of the monomers (Figs. 1a (lane L), 2a, and 6 (a-c)). Pulse-chase experiments revealed that the ␣3 chain signal seemed to decay faster than the other two subunits (Fig. 3b). We postulate that the ␣3 chain is limiting for heterotrimer assembly and that the association of the ␣3 chain with ␤3␥2 heterodimers occurs rapidly to yield complete heterotrimers. A precedent in laminin-1 from murine teratocarcinoma cell lines has been shown in which the ␣1 chain is limiting for assembly at the protein level, in accordance with its low mRNA levels (24,25). One possible rationale for the synthesis of the ␣3 chain being limiting in the synthesis of laminin-5 is that the cell uses ␣ chain synthesis as a determining step in controlling the type of laminin to be assembled. Thus, expression levels of either ␣1 or ␣2 may determine whether laminin-1 (␣1␤1␥1) or merosin (␣2␤1␥1) is to be assembled or whether s-laminin (␣1␤2␥1) or s-merosin (␣2␤2␥1) is assembled. Although no laminin variant containing ␤3␥2 other than laminin-5 (␣3␤3␥2) has yet been described, the possibility of additional laminin variants seems likely. One circumstance in which laminin-5 production would be crucial is in activated keratinocytes repopulating a wound bed. Keratinocytes in tissue culture display markers for the activated phenotype as well, and the ␣3 chain pool typically exceeds the other monomer pools in these cells (data not shown).
Following assembly of the heterotrimer, processing of two out of the three chains occurs, ␣3 and ␥2. The cytoplasmic form of laminin-5 was previously identified as a 460-kDa precursor that contains unprocessed forms of each chain (2). Processing occurs after secretion, with two predominant forms of 440 and 400 kDa present in the medium fraction. Processing of the ␣3 chain to 165 kDa has already occurred in both of these forms. However, in our pulse-chase experiments, faint bands of 200 and 180 kDa were also visible (Fig. 3d, 3-, 6-, and 24-h chase samples). The 180-kDa species may be an intermediate in the processing of the 200-kDa ␣3 chain to its final processed form of 165 kDa. Since these bands were minor components, they presumably belong to a minor fraction of heterotrimers secreted into the medium that were not yet processed at the moment of sampling. Resolution of a small amount of the unprocessed heterotrimer, and the presumed processing inter-mediate, may not be readily resolvable from the major form of 440 kDa in these gels.
We resolved the 440-and 400-kDa forms of heterotrimeric laminin-5 in the medium fraction into the component chains in our SCC-25 cell model system, as was previously done in keratinocytes (2) (Fig. 1c, lanes L, ␤, and ␥). The larger species consists of 165-kDa ␣3, 145-kDa ␤3, and 155-kDa ␥2 chains, while the smaller one consists of 165-kDa ␣3, 145-kDa ␤3, and 105-kDa ␥2 chains (Fig. 2b). Furthermore, in pulse-chase experiments of the nonreduced medium samples (Fig. 3c), the 440-kDa form of laminin-5 first appeared after the 1.5-h chase and was clearly evident after the 3.0-h chase, whereas the 400-kDa form appeared later, only after a 3-h chase, clearly visible after a 6-h chase. Consistent with the 440-and 400-kDa forms being composed of the same chains but with processed ␥2 (105 kDa) substituted for unprocessed ␥2, the processed form of ␥2 (105 kDa) appears with the same kinetics as the 400-kDa form. These results support a precursor-product relationship between the 440-and 400-kDa forms, due to processing of the ␥2 chain from 155 to 105 kDa. Dimers or monomers were not detected in SCC-25 medium, consistent with a lack of secretion in the absence of heterotrimer assembly, although these forms might be present at steady-state concentrations below the detection limit of the assay. A 250-kDa protein was coprecipitated with laminin-5. Since the intensity of this signal was reduced by preincubation with gelatin, this protein may be fibronectin. Coprecipitation of fibronectin with laminin-5 was previously reported in normal and HJEB human keratinocyte systems (26).
Multimers of individual chains, such as a ␤1 chain multimer as reported for mouse laminin-1 (7), were not detected in laminin-5 from SCC-25 cells (Figs. 1a (lanes L, ␤, and ␥) and 2a). It is possible that these forms might be present at steady-state concentrations below the detection limits of the assay or are immunologically unreactive. The latter possibility is less likely since ␤ and ␥ multimers were detected in transfected 293 cells by the same antibodies. Multimer bands have previously been noted when other proteins have been expressed in heterologous systems, for example the expression of human fibrinogens ␣, ␤, and ␥ in baby hamster kidney cells (22). In a study on the influenza hemagglutinin protein, only properly folded multimeric proteins were transported out of the rough endoplasmic reticulum, whereas incompletely folded proteins either accumulated or were degraded in the endoplasmic reticulum (27). One characteristic of the 293 cells used to conduct our transfection experiments may be a reduced capacity to deal with improperly folded proteins.
We have shown that the assembly of recombinant ␤3 and ␥2 chains is achieved without the presence of the ␣3 chain (Fig. 4a,  lanes 5 and 11). In a further analysis of the cotransfection products on two-dimensional SDS gels, cells cotransfected with ␤3 and ␥2 expression vectors produced both monomers (Fig. 5,  a and b, black and white asterisks) and the ␤␥ heterodimer (Fig.  5, a and b, black and white arrows). When anti-␤3 antibody was used, the major bands were ␤ monomers (145 kDa) and ␤␥ dimers (300 kDa). The coexpression of ␤ and ␥ leads mainly to the formation of a ␤␥ dimer held together by disulfide bond(s) as well as free ␤ chain. When anti-␥2 antibody was used, a ␤␥ dimer signal was also detected. Since this antibody does not recognize the ␤␥ heteromultimer (Fig. 5a, black and white  arrows), heteromultimerization might obstruct binding to antibody recognition sites. We conclude that the ␣3 chain is not necessary for ␤␥ heterodimer formation, but that it is an intermediate in heterotrimer formation in SCC-25 cells. Considering that laminin-1 assembly probably proceeds via a ␤␥ dimer, and from our results with laminin-5, it seems a reasonable prediction that other laminin isoforms also will be shown to assemble via a ␤␥ dimer intermediate.
The pulse-chase study demonstrated that newly synthesized laminin-5 subunits appear in the cells immediately following a 10-min biosynthetic pulse with [ 35 S]methionine and [ 35 S]cystine. Since other related proteins are known to undergo addition of N-linked glycochains after a 1-h chase, as reported for laminin-1 of human choriocarcinoma cells (3) and mouse embryonic carcinoma F9 cells (4), we expected to observe an increased mobility of the laminin-5 chains on SDS-PAGE due to the lack of glycosylation of these intracellular precursors. However, precursors with altered electrophoretic mobility were not detected on two-dimensional SDS-PAGE (Fig. 2a) and in the pulse-chase study (Fig. 3b). These results suggest that the high mannose chains on the subunits are not processed into complex forms.
To further investigate the role of glycosylation in laminin-5 chain assembly, SCC-25 cells were treated with tunicamycin to inhibit the addition of asparagine-linked carbohydrates, and cell and medium fractions were analyzed by SDS-PAGE and two-dimensional SDS-PAGE. The heterotrimer, heterodimer, and monomers were present in the cell fraction, but had slightly increased mobility on the gels due to the inhibition of glycosylation. The presence of the heterotrimer suggested that the assembly of unglycosylated subunits also proceeds through a ␤3␥2 heterodimer to the ␣3␤3␥2 heterotrimer. Since protein disulfide-isomerase is present at the luminal side of the rough endoplasmic reticulum (28), disulfide bond formation between laminin-5 subunits is expected to be completed before they leave the rough endoplasmic reticulum. Additionally, it is well established for N-glycosylated proteins that transfer of high mannose-type oligosaccharide side chains en bloc proceeds cotranslationally (29). Dimer and trimer formation in the presence of tunicamycin demonstrates that the inhibition of this oligosaccharide transfer by tunicamycin does not affect subsequent disulfide bond formation. However, high mannose oligosaccharide chains can have a profound effect on the stability of proteins (30,31), as has also been shown for laminin-1 (3). The amount of laminin-5 in the medium fraction was extremely small, so we postulate that N-glycosylation protects laminin-5 polypeptides from nonspecific proteolytic degradation. Fig. 7 presents a summary diagram of our conclusions concerning post-transcriptional assembly and glycosylation of laminin-5 subunits. These include the notions that 1) assembly proceeds through a ␤3␥2 heterodimer to the ␣3␤3␥2 heterotrimer; 2) synthesis of ␣3 polypeptides is the rate-limiting step for assembly; 3) assembly of the subunits into a heterotrimer is required for secretion; and 4) N-linked oligosaccharide chains are not necessary for subunit assembly.
As reported (24), HJEB patients have impaired expression of laminin-5, which is most often a consequence of the defective synthesis of one of its subunits. The disease is genetically heterogeneous, even considering the HJEB patient population with laminin-5 defects, since any one or combination of chain defects is possible within the heterotrimer. There is a range of possible mutations in laminin-5 that may have effects on assembly, stability, or secretion of the protein, and consequently, individual patients might be expected to display different clinical phenotypes. Chain assembly of laminins is mediated through the formation of triple-stranded ␣-helical coiled-coils, known to be among the longest coiled-coil domains. In fibrinogen, another triple-stranded ␣-fibrous protein, the chain specificity is determined by interactions between residues adjacent to the hydrophobic interaction edges. Protein sequence data have been used to calculate ionic interaction scores between heptad repeat regions in laminin-1 chains that were in good agreement with experimental observations and that may allow predictions of the stability of distinct laminin isoforms (32,33). Now that we have noted the appearance of a ␤3␥2 heterodimer as a likely assembly intermediate for laminin-5, we would expect some of the mutations contained in HJEB patients to effect ␤3␥2 dimerization, while another class of mutations might interfere with ␣3 association with the heterodimer to form a functional heterotrimer. We are currently correlating HJEB patient phenotypes with characterization of molecular defects to ascertain the extent to which we are able to use this knowledge to predict clinical features and ultimately the course of the disease.