Structural and Functional Analysis of the Recombinant G Domain of the Laminin α4 Chain and Its Proteolytic Processing in Tissues*

The C-terminal G domains of laminin α chains have been implicated in various cellular and other interactions. The G domain of the α4 chain was now produced in transfected mammalian cells as two tandem arrays of LG modules, α4LG1–3 and α4LG4–5. The recombinant fragments were shown to fold into globular structures and could be distinguished by specific antibodies. Both fragments were able to bind to heparin, sulfatides, and the microfibrillar fibulin-1 and fibulin-2. They were, however, poor substrates for cell adhesion and had only a low affinity for the α-dystroglycan receptor when compared with the G domains of the laminin α1 and α2 chains. Yet antibodies to α4LG1–3 but not to α4LG4–5 clearly inhibited α6β1 integrin-mediated cell adhesion to laminin-8, indicating the participation of α4LG1–3 in a cell-adhesive structure of higher complexity. Proteolytic processing within a link region between the α4LG3 and α4LG4 modules was shown to occur during recombinant production and in endothelial and Schwann cell culture. Cleavage could be attributed to three different peptide bonds and is accompanied by the release of the α4LG4–5 segment. Immunohistology demonstrated abundant staining of α4LG1–3 in vessel walls, adipose, and perineural tissue. No significant staining was found for α4LG4–5, indicating their loss from tissues. Immunogold staining demonstrated an association of the α4 chain primarily with microfibrillar regions rather than with basement membranes, while laminin α2 chains appear primarily associated with various basement membranes.

The protein family of laminins consists of at least 12 different isoforms, which are mainly localized in basement membranes. They are involved in major biological functions such as interactions with cellular receptors and the formation of networks that are intermingled with and bound to networks of collagen type IV (1,2). Most of these heterotrimeric isoforms consist of ␤1/␤2 and ␥1 chains but differ in their ␣ chains, ␣1 to ␣5. The ␣4 chain (200 kDa) is the shortest variant known so far and is present in laminin-8 (␣4␤1␥1) and laminin-9 (␣4␤2␥1) (3)(4)(5). The existence of such relatively small laminins was originally indicated from biosynthetic studies with endothelial and adipose cells (6,7), but their molecular nature was only understood after the complete human (8,9) and mouse (4, 10, 11) ␣4 chain sequences became available. The domain structure of the ␣4 chain (1816 residues) predicted a small Nterminal region contributing a truncated short arm structure, a coiled-coil domain II/I used for chain association, and a large C-terminal G domain. This prediction was confirmed by electron microscopy of laminin-8 and -9, which lacked one of the three short arm structures found in other laminins (3,4).
Northern and in situ hybridization demonstrated a moderate to strong expression of the ␣4 chain in heart, lung, skeletal muscle, and skin, while some other tissues were negative (4, 8 -11). The chain was also expressed at midgestation stages of mouse development (4,11) and in various endothelial and adipocyte cell lines (3,4). Antibodies raised against fusion proteins of the ␣4 chain were useful in showing the extracellular deposition of the corresponding laminins by immunohistology (5,11). This demonstrated a distinct localization in striated muscle, perineurium, capillaries, and some mesenchymal regions but only a low abundance in epithelial basement membrane zones. This suggested that ␣4 chains are a distinct component of subendothelial regions and that they may have an adhesive function for endothelial cells (11). It was also speculated that they may promote angiogenesis (3,7).
Specific binding functions have not yet been examined for the ␣4 chain G domain, although such functions are shared by the G domains of all other laminin ␣ chains (2). These G domains consist of a tandem array of five LG modules, LG1 to LG5, each of about 200 amino acid residues. Previous data for ␣1 and ␣2 chains showed the involvement of their G domains in integrinmediated cell adhesion and binding to heparin, sulfatides, and the ␣-dystroglycan receptor (2,(12)(13)(14)(15). Some of the binding epitopes could be mapped by site-directed mutagenesis to the laminin ␣1LG4 and ␣2LG5 modules and showed a considerable overlap (13,14). The recent elucidation of the crystal structure of ␣2LG5 (16) was instrumental in understanding the spatial organization of these epitopes. Furthermore, a recombinant fragment corresponding to laminin ␣2LG1-5 was shown to promote the attachment of Mycobacteria leprae to Schwann cells (17,18), indicating that LG modules are also likely to be involved in pathological processes.
Based on our previous experience with the recombinant production of LG modules of laminin ␣1 and ␣2 chains in mammalian cells (13,19), we have now prepared the tandem arrays ␣4LG1-3 and ␣4LG4 -5 for the mouse laminin ␣4 chain. These fragments had a strong affinity for heparin but no activity or only little activity in cell adhesion and the binding of ␣-dystroglycan. Furthermore, the data indicated a substantial absence of the ␣4LG4 -5 structure from tissues due to proteolytic processing.
Construction of Expression Vectors-Mouse laminin ␣4 chain cDNA clone M16 (10) was used as a template to amplify the sequence encoding the ␣4LG4 -5 modules (residues 1428 -1816) by polymerase chain reaction with Vent polymerase (New England Biolabs) following the manufacturer's instructions. The primers used were GTCAGCTAGCGGAT-GCGCCTTCATGGG for the 5Ј-end and GTCACTCGAGTCAGGCTGT-GGGACAGGA for the 3Ј-end. In addition to the coding sequences, these primers introduced a stop codon and single NheI and XhoI restriction sites in order to allow in-frame insertion of the cDNA distal to the BM-40 signal peptide sequence in the episomal expression vector pCEP/pu (29). Clones M47 and M16 (10) were used for the preparation of the laminin ␣4LG1-3 construct (residues 827-1427) in two steps. The primer pairs GTCAGCTAGCAGTCTCCATGATGTTTG and ACGTGC-CGTCTGTCCAC (for M47) and GTGGACAGACGGCACGT and GTCA-CTCGAGCTACTTACTCTTCTCTCCC (for M16) were used for amplification, and the two polymerase chain reaction products were then fused by overlap extension. The final polymerase chain reaction-derived construct contained the same restriction sites and a stop codon as the construct described above. Both were initially ligated into plasmid pUC18 (Amersham Pharmacia Biotech) for sequence verification on a 373A automated sequencer (Applied Biosystems). They were then released by NheI and XhoI digestion and ligated into plasmid pCEP/Pu (29).
Expression and Purification of Recombinant Proteins-Human embryonic kidney cells that constitutively express the EBNA-1 protein from Epstein-Barr virus (293 EBNA; Invitrogen) were transfected with the episomal expression vectors (29), and transfected cells were selected with 0.5 g/ml puromycin (Sigma) and 250 g/ml G418 (Life Technologies, Inc.). They were washed extensively with phosphate-buffered saline (pH 7.2) to remove residual serum proteins and grown in serumfree Dulbecco's modified Eagle's medium/F-12 medium (Life Technologies) for 2 days, after which medium was harvested and new serum-free medium was added for another 2 days. Conditioned serum-free medium (1l) was dialyzed against 0.05 M Tris-HCl, pH 7.4, containing 0.5 mM phenylmethylsulfonyl fluoride (Serva) and 0.5 mM N-ethylmaleimide (Merck). It was then passed over a 2 ϫ 30-cm heparin-Sepharose column, which was equilibrated in the same buffer and eluted with a linear NaCl gradient (0 -0.6 M NaCl, 500 ml). Recombinant proteins were further purified on a Superose 12 column (HR16/50; Amersham Pharmacia Biotech) equilibrated in 0.2 M ammonium acetate, pH 6.8, lyophilized, and redissolved in 0.2 M NH 4 HCO 3 .
Purification of Laminin Proteins from Culture Medium-Conditioned serum-free medium (0.5-1l) was harvested from eEnd.2 and rat Schwannoma RN22 cells. After the addition of protease inhibitors (0.05 mM Pefabloc, 1 mM EDTA, 0.5 mM N-ethylmaleimide), medium was dialyzed against 0.1 M NaCl, 0.05 M Tris-HCl, pH 7.4, and passed over a 5-ml heparin HiTrap column (Amersham Pharmacia Biotech) equilibrated in the same buffer. Bound proteins were eluted with a 0.1-0.6 M NaCl gradient (60 ml). Concentrated pools were subsequently passed over a Superose 12 column (HR 10/30) in 0.2 M ammonium acetate, pH 6.8, and analyzed by immunoblotting and by SDS-gel electrophoresis using Coomassie Blue staining.
Analytical Methods-Protein and hexosamine concentrations were determined on a Biotronik LC3000 analyzer after hydrolysis (16 h, 110°C) with 6 or 3 M HCl, respectively. Edman degradation was performed with 473A or Procise sequencers, following the manufacturer's instructions. Electrophoresis in SDS-polyacrylamide gradient gels followed standard protocols. Circular dichroism spectra were recorded using a J-175 spectropolarimeter (Jasco Labor) and evaluated as described previously (30). Rotary shadowing electron microscopy was carried out according to established procedures (31).
Ligand Binding Assays-A 1-ml heparin-HiTrap column (Amersham Pharmacia Biotech) in 0.05 M Tris-HCl, pH 7.4, was used to determine the NaCl concentration required to displace bound ligands from the column with a precision of Ϯ0.01 M NaCl (13,14). Solid-phase binding assays were carried out with various proteins (5 g/ml) and heparinalbumin conjugate (10 g/ml) adsorbed onto the plastic surface of microtiter wells at 4°C following a previous procedure (32) with some modifications (14). Coating with sulfatides dissolved in methanol (0.2 mg/ml; 50 l) was performed by drying overnight at room temperature. 1 mM CaCl 2 and MgCl 2 were added to the buffer in the assays with ␣-dystroglycan. Binding of soluble ␣4LG1-3 and ␣4LG4 -5 was detected by specific antisera (see below). Surface plasmon resonance assays were performed with BIAcore 1000 instrumentation (BIAcore AB) using proteins coupled through carbodiimide to CM-5 sensor chips (research grade). Binding assays were carried out in neutral buffer containing 2 mM CaCl 2 under controlled conditions to prevent mass transport problems (33). Kinetic constants were calculated by nonlinear fitting of association and dissociation curves according to a 1:1 model following the manufacturer's instructions (BIAevaluation software version 3.0).
Cell Adhesion Assays-Cell attachment to plastic-coated laminin fragments was detected by rigorous washing followed by staining with 0.1% crystal violet and colorimetry according to established protocols (34). Collagen IV, fibronectin, and laminin-1 were used as positive controls (26). Adhesion to bovine serum albumin, which was used for the blocking of coated wells, was negligible. Adhesion-blocking monoclonal antibodies against ␣ 6 (GoH3) and ␤ 1 (AIIB2) integrin subunits were kindly provided by A. Sonnenberg and C. H. Damsky. They were used together with substrate-specific antibodies in inhibition assays (34).
Immunological Assays-Rabbit antisera were generated against the two ␣4 chain fragments by two injections of 0.2 mg in complete Freund's adjuvant, and antibodies were affinity-purified (35). Rabbit antibodies against mouse laminin fragments ␣2LG1-3 and ␣2LG4 -5 have been previously described (19). Enzyme-linked immunosorbent assay titrations followed standard protocols. Immunoblotting followed a previously used procedure (36).
Immunohistochemistry-Paraffin sections of adult NMRI mice were deparaffinized, rehydrated, and incubated (10 min) with protease XXIV (Sigma) to block endogenous peroxidase. They were then exposed for 1 h at room temperature to affinity-purified rabbit antibodies against ␣2 chain (19) and ␣4 chain fragments diluted to 5-7 g/ml. Peroxidase anti-peroxidase staining and counterstaining with hematoxylin followed a previously described procedure (37). Negative controls were carried out with normal rabbit serum diluted 1:100. Frozen tissue sections were used for indirect immunofluorescence (36).
Tissue sections on nickel grids were used for indirect immunogold staining (38). They were incubated for 1 h at room temperature with affinity-purified antibodies (10 -15 g/ml), rinsed, and incubated for 20 min with affinity-purified goat anti-rabbit IgG (Medac, Hamburg) coupled to 16-nm gold particles diluted 1:300. Sections were rinsed with water, stained with uranyl acetate (15 min) and lead citrate (5 min), and then examined with a Zeiss EM 109 electron microscope. Controls with antibody-coated or -uncoated gold particles were all negative.

Recombinant Production of Two Fragments Comprising the G Domain of the Laminin ␣4
Chain-The G domain of the mouse laminin ␣4 chain was prepared in the form of two recombinant fragments, ␣4LG1-3 (residues 827-1427) and ␣4LG4 -5 (residues 1428 -1816) following a previous strategy used for the laminin ␣2 chain (19). The boundaries chosen were outside the predicted ␤ sandwich structure of the LG modules (16), and the border between the two fragments was placed in the center of a long link region (residues 1398 -1460). Both fragments were produced and were obtained in good yields (1-2 g/ml) after purification. Because of their strong heparin affin-ity, they could be readily purified by a two-step chromatographic procedure, as shown by electrophoresis (Fig. 1).
Fragment ␣4LG1-3 migrated as a band of 67 kDa and showed a single N-terminal sequence APLAVSM, where APLA is derived from the foreign signal peptide region. Fragment ␣4LG4 -5 could be separated by electrophoresis into two bands of 43-44 kDa. Edman degradation of the upper band demonstrated the expected APLADAPXWD sequence. Two sequences, XKFLEQKA and XEQKAP, which were identified for the lower band, represent starting positions of 1437 and 1440, respectively, indicating proteolytic trimming within the linker region. Hexosamine analysis of ␣4LG1-3 demonstrated 7 residues of glucosamine but no galactosamine, in agreement with the presence of four potential N-glycosylation sites in the sequence (10,11). No hexosamine could be detected in fragment ␣4LG4 -5, which lacks N-glycosylation sites.
Both fragments were folded into compact globular structures, as shown by electron microscopy (Fig. 2). They thus had the same shape as previously shown for analogous tandem arrays of LG modules derived from laminin ␣1 and ␣2 chains (13,19). Circular dichroism spectra of ␣4LG1-3 and ␣4LG4 -5 (data not shown) were nearly identical to that previously published for the proteolytic fragment E3 of laminin-1, which corresponds to ␣1LG4 -5 (39). They showed a minimum at 210 -215 nm ( ϭ Ϫ5500 to 8000 degrees⅐cm 2 ⅐dmol Ϫ1 ), indicating a content of 47-60% ␤ strands and ␤ turns. Together, the data demonstrated that both recombinant fragments of the laminin ␣4 chain were properly folded.
Binding to Sulfated Ligands and Extracellular Matrix Proteins-Binding of laminin fragment E3 to heparin (39) and sulfatides (40) were the first activities assigned to laminin LG modules and subsequently confirmed with various other recombinant LG fragments (13,14). Fragments ␣4LG1-3 and ␣4LG4 -5 were similar in this context (Table I). They bound quantitatively to an analytical heparin affinity column and needed 0.27 and 0.34 M NaCl, respectively, for displacement. This indicated that they are potential ligands for heparin/ heparan sulfate at physiological ionic strength, as found before for recombinant LG fragments of the laminin ␣1 and ␣2 chains (13,14). Their binding activities for a heparin-albumin conjugate and for sulfatides in solid phase assays were, however, distinctly lower than ␣2 chain fragments ( Table I). As shown previously (13), recombinant fragment ␣1LG4 -5 is also a stronger ligand in both solid-phase assays (half-maximal binding at 4 -6 nM).
The basement membrane proteins fibulin-1, fibulin-2, and nidogen-2 were used as ligands for ␣4LG1-3 and ␣4LG4 -5 in surface plasmon resonance assays in order to compare their binding activities with those previously determined for similar ␣2 chain fragments (Table II). The ␣4 fragments bound to both fibulins, although the affinities differed 2-10-fold from those of the ␣2 chain fragments. Nidogen-2 was a poor ligand for ␣4LG4 -5 and did not bind to ␣4LG1-3. Several other proteins (nidogen-1, perlecan, BM-40, and collagens I and IV) were also tested with ␣4LG1-3 and ␣4LG4 -5 in solid-phase assays but showed no binding or only marginal binding, which did not reach plateau levels, up to a concentration of 1 M for the soluble ligands.
Interactions with Cellular Receptors-LG modules have previously been shown to be good candidates for binding to ␣-dystroglycan, which is an important receptor in many cell types (41). Immobilized ␣-dystroglycan was therefore used in solidphase assays to compare the binding of ␣4LG1-3, ␣4LG4 -5, and ␣2LG1-3 fragments (Fig. 3). This demonstrated a strong binding of ␣2LG1-3 as shown before (14). The two ␣4 chain fragments, however, were only poor ligands, which did not reach plateau levels up to a concentration of 1 M. This indicated a 30 -100-fold lower binding activity than ␣2LG1-3.
Previous studies have shown that ␣2LG1-3 but not  ␣2LG4 -5 strongly promotes ␤ 1 integrin-mediated attachment and spreading of several cell lines (15). Three of these cell lines, Rugli glioma (Fig. 4), RN22 Schwannoma, and epithelial HBL100 cells, showed no distinct binding to fragment ␣4LG1-3. Because of the localization of laminin ␣4 chain in various vessel walls (5,11), it was of particular interest to examine endothelial and smooth muscle cells in these assays. Pulmonary artery smooth muscle cells (Fig. 4) and aortic endothelial cells attached rather weakly to ␣4LG1-3 and ␣2LG1-3 and not at all to ␣4LG4 -5. Four further endothelial cell lines (see "Materials and Methods") showed no significant binding to the three substrates tested. By contrast, an embryonic endothelial cell line (27) bound strongly to ␣2LG1-3, exceeding the level of binding of Rugli cells, but did not attach to ␣4LG1-3 or ␣4LG4 -5 substrates (data not shown). Fibronectin and collagen IV were used as positive controls in the assays and were strongly adhesive for all cells examined, in agreement with previous observations (26). Recombinant laminin-8 (␣4␤1␥1), however, was recently shown to promote adhesion of HT1080 cells by binding to ␣ 6 ␤ 1 integrin (42). We could now show the same for laminin-8 from RN22 cells by using blocking monoclonal antibodies. This interaction could also be inhibited in a dose-dependent manner by incubating the substrate with affinity-purified antibodies (see below) against ␣4LG1-3 but not against ␣4LG4 -5 (Table   III). Together, the data indicate contributions of ␣4LG1-3 to cell adhesion but only in the context of an entire laminin structure.
Immunological Analyses of Cells and Tissues-Since the recombinant data indicated a possible proteolytic processing of the G domain of the laminin ␣4 chain in situ, we generated rabbit antisera against fragments ␣4LG1-3 and ␣4LG4 -5. Antibodies were purified by affinity chromatography on the antigen used for immunization. As shown in Fig. 5, the antibodies against ␣4LG1-3 did not cross-react substantially with ␣4LG4 -5, ␣1LG1-3, and ␣2LG1-3. A similar high specificity was also observed for the antibodies against ␣4LG4 -5. This specificity was confirmed by immunoblotting of the ␣4 chain fragments (Fig. 6, A and B, lanes 1 and 5).
The antibodies against the two different ␣4 chain epitopes both showed distinct reactions with various cultured cells and their conditioned media by immunofluorescence or immunoblots. The two antibodies showed quite different staining patterns in reduced immunoblots of culture medium from the mouse endothelial cell line eEnd.2, the rat Schwannoma RN22 cells, and endothelial cells (HUVEC) from the human umbilical vein (Fig. 6, A and B, lanes 2-4). Antibodies against ␣4LG4 -5 reacted mainly with 2-3 bands of about 43-45 kDa but also with bands of about 210 kDa. Antibodies against ␣4LG1-3, however, primarily bound to bands in the range 180 -210 kDa with only little reaction with smaller bands. Together, the data indicated a substantial release of the ␣4LG4 -5 structure by proteolytic processing but also a certain variability in the cleavage sites.
RN22 cell medium was used to separate individual ␣4 chain components by heparin affinity and molecular sieve chromatography, which was monitored by immunoblotting. This allowed a partial separation of the 200-kDa components from the 45-kDa bands, which eluted later from the heparin column. A final separation of the 45-kDa fragments was then achieved on a Superose 12 column equilibrated in neutral buffer. Electrophoresis of this material demonstrated a 43/45-kDa doublet band, which was, however, still contaminated with some other proteins (Fig. 1, lane 5). Edman degradation after blotting the doublet demonstrated the sequences XEKSKDAPSW (upper band) starting at position 1423 of the ␣4 chain and LK-FLEXKAP (lower band) starting at position 1437. A similar separation could be achieved for eEnd.2 medium, but the yields  were insufficient for sequencing. Since all separations were performed under nondissociating conditions, it indicates that, once released, the ␣4LG4 -5 structure does not stay associated with the remaining laminin.
Immunohistology was used to determine the mouse tissue localization of ␣4LG1-3 and ␣4LG4 -5 at the light and electron microscopical level and to compare it with that of correspond-ing laminin ␣2 chain fragments. Affinity-purified antibodies against ␣4LG1-3 showed a distinct staining (peroxidase technique) of capillary walls in heart (Fig. 7A) and skeletal muscle (Fig. 7F) but failed to react significantly with basement membrane zones (endomysium) around the muscle cells. No staining was observed with antibodies against ␣4LG4 -5 as shown for heart muscle (Fig. 7B). Comparable staining patterns of both endomysium and capillaries could be obtained with antibodies against ␣2LG1-3 (Fig. 7, C and E) and ␣2LG4 -5 (Fig.  7D). Further strong reactions for ␣4LG1-3 were detected in basement membrane zones around smooth muscle cells of skin blood vessels, in bronchial regions, in alveolar septa, around the perineurium, and in the tunica media and around adventitial adipocytes of aorta. None of these regions reacted with antibodies against ␣4LG4 -5, in contrast to ␣2LG1-3and ␣2LG4 -5-specific antibodies, which produced indistinguishable staining patterns in all tissues examined. Staining for ␣4LG1-3 but not for ␣4LG4 -5 was also confirmed by indirect immunofluorescence on various frozen tissue sections (data not shown).
Immunogold staining with antibodies to ␣2LG1-3 and ␣4LG1-3 was used in order to distinguish laminin ␣2 and ␣4 chains at the ultrastructural level. The ␣2 chain could be clearly detected within basement membranes around skeletal muscle cells and adjacent to endothelial cells and pericytes of capillaries and small arterioles (Fig. 8, A and B). In heart muscle, however, no basement membrane staining was found around cardiomyocytes, but staining occurred in deeper microfibrillar layers of the endomysium (Fig. 8C). By contrast, the ␣4 chain was not a basement membrane component of either endothelial or muscle cells but was instead located in the adjacent interstitial region of skeletal muscle (Fig. 8D) and heart muscle (Fig. 8, E and F). DISCUSSION The recombinant production of the mouse laminin ␣4 chain domain G in the form of two tandem arrays, ␣4LG1-3 and ␣LG4 -5, as described here, has set the stage for several functional and biological studies. Electron microscopy and circular dichroism spectroscopy demonstrated that they were properly folded, as shown before for analogous fragments from the laminin ␣2 chain (19), which made the ␣4 modules suitable for ligand binding studies. A limited proteolytic processing of recombinant ␣4LG4 -5 led us to investigate whether similar processing may occur in cell cultures and tissues. Proteolytic processing has been previously identified within the ␣2LG3 module of the laminin ␣2 chain (19) and was predicted to occur between the ␣3LG3 and ␣3LG4 modules of the laminin ␣3 chain (43). No cleavage has yet been reported for the G domain of the laminin ␣1 chain.
When compared with the laminin ␣2 chain fragments, both recombinant ␣4 fragments bound with a similar strength in heparin affinity chromatography but showed a more moderate interaction in solid phase assays with heparin and sulfatides. Such binding could be important for cellular interactions and, as shown for the laminin ␣1LG4 module (44,45), also for binding to the heparan sulfate chains of the extracellular proteoglycan perlecan. The heparin/sulfatide binding epitopes have been mapped by site-directed mutagenesis to a few basic residues in the laminin ␣1LG4 (13) and ␣2LG5 (14) modules. Furthermore, the crystal structure of ␣2LG5 (16) demonstrated that they are localized in short loops between ␤ strands F/G and H/I for ␣1LG4 and between ␤ strands H/I and L/M for ␣2LG5, which are in close proximity to the surface of LG modules. The basic character of these loops is maintained in all five of the LG modules of mouse and human laminin ␣4 chains (4, 8 -11). Their role in binding can now be examined by appro-  priate mutants of recombinant ␣4LG1-3 and ␣4LG4 -5 fragments.
The LG modules of the laminin ␣2 chain were previously shown to bind to fibulin-1, fibulin-2, and nidogen-2 (14), interactions that could be important for the supramolecular organization of extracellular structures. Similar interactions, with some differences in binding affinities, could now be demonstrated for ␣4LG1-3 and ␣4LG4 -5, suggesting that this property could also be shared by other laminin ␣ chains. This is supported by previous studies, which showed binding of a laminin fragment E3 (␣1LG4 -5) to fibulin-1 (46) and recent observations on the binding of ␣1LG1-3 to fibulin-2. 1 The fibulins and nidogen-2 are known to occur in basement membranes but are also associated with fibrillin and fibronectin microfibrils and elastic sheets (36,(47)(48)(49).

FIG. 8. Immunogold localization of ␣2LG1-3 (A-C) and ␣4LG1-3 (D-F) structures in adult mouse skeletal muscle and heart tissues.
A and B, staining of soleus muscle shows localization of laminin ␣2 chain in basement membranes (asterisks) around a myocyte (my) and along endothelial cells (en, arrows), and along endothelial cells (en, asterisks) and around a pericyte (pe, arrows) of a small arteriole. l, vessel lumen. C, heart muscle laminin ␣2 chain is found in the interstitial matrix of the endomysium but not in the basement membrane (asterisks) around the cardiomyocyte (my). D, staining of soleus muscle shows laminin ␣4 chain in the interstitial matrix adjacent to a capillary but not in the endothelial cell (en) basement membrane (arrows). E and F, a similar staining of heart muscle reveals labeling of the interstitial matrix next to a small arteriole (E) and a capillary (F) but not in the basement membranes (asterisks) adjacent to endothelial cells (en) and cardiomyocytes (my). D-F also show an erythrocyte (ery) within the capillary lumen (l) and a pericyte (pe). Bars, 0.32 m (A and C-F) and 0.43 m (B).
LG modules in situ now needs to be examined by immunogold colocalization studies.
Laminin LG modules are also important ligands for cellular receptors, including several integrins and ␣-dystroglycan (12,41). Here we show a rather low binding of ␣4LG1-3 and ␣4LG4 -5 to ␣-dystroglycan when compared with LG modules derived from perlecan and laminin ␣1 and ␣2 chains (13,14). Studies with laminin ␣1LG4 demonstrated that ␣-dystroglycan binding depends on residues involved in heparin/sulfatide binding as well as several other basic amino acids (13) that are located more distantly in loops between ␤ strands J/K, K/L, and M/N (16). These latter regions are not very well conserved in the LG modules of the laminin ␣4 chains, which may explain the low binding activity. The laminin ␣2LG1-3 but not the ␣2LG4 -5 fragment was a strongly cell-adhesive substrate, mediated by interactions with ␣ 3 ␤ 1 and ␣ 6 ␤ 1 integrins (15). Fragment ␣4LG1-3, however, was a poor adhesive substrate for several standard tumor cells and endothelial cells. However, the laminin ␣1LG1-3 structure needs to be associated with the adjacent rod domain of the long arm to express strong binding activity for ␣ 6 ␤ 1 integrin (50,51). Laminin-8 was in fact recently shown to bind cells via the ␣ 6 ␤ 1 integrin (42,52), which could be confirmed in the present study. This interaction was furthermore specifically inhibited by antibodies against ␣4LG1-3 but not by antibodies against ␣4LG4 -5 (Table III). This suggests, like for ␣1LG1-3, that interactions between ␣4LG1-3 and the rod are required for the expression of a strong cell-adhesive epitope.
Proteolytic processing of the LG region of the laminin ␣4 chain was confirmed with cultured endothelial and Schwannoma cells. It occurred in a 65-residue link region between the ␣4LG3 and ␣4LG4 modules and included three different cleavage sites (Fig. 9). The principal fragments released showed a limited size heterogeneity and included C-terminal fragments of 43-45 kDa and N-terminal fragments of 180 -210 kDa. The latter correspond to the 180 -200-kDa ␣4 chains previously detected in embryo extracts (5), leiomyosarcoma cells (11), and platelet laminin-8 (52). This indicates at least a partial release of ␣4LG4 -5, which, however, was not identified in the previous studies. It is also noteworthy that the identified cleavage sites are not entirely conserved in the mouse and human ␣4 chains (Fig. 9). Together with the multiple cleavage sites, this suggests that several types of proteases could be involved in processing. As a consequence, the ␣4LG4 -5 entity no longer remains associated with the parental laminin. This is different from the processing of the laminin ␣2 chain, where proteolysis occurs at a single Arg-Gln bond at the C-terminal end of a furin-type cleavage sequence of the ␣2LG3 module (15,19). This cleavage is not accompanied by dissociation, probably due to the fact that cleavage occurs in a longer insert in the loop between ␤ strands D and E of the LG module and should therefore not disrupt the ␤ sandwich (16).
A conserved feature of the link region is an odd cysteine close to its C-terminal end (Fig. 9). A recent crystal structure of the ␣2LG4 -5 tandem array (53) demonstrated that this cysteine forms an intermodular disulfide bridge to a cysteine in the short ␣-helix of the ␣2LG5 module. Furthermore, about 15 C-terminal residues of the link form an extended interface contact region, which forces a distinct and stable topological orientation of the modules relative to each other. Based on sequence comparisons and the fact that processing occurs on the N-terminal side of the interacting link region, it is likely that the same tertiary structure should be present in ␣4LG4 -5 and in all other laminin ␣ chains. This may also be the case for other LG modules present in protein S and the receptor kinase ligand Gas6 (53).
Immunolocalizations with antibodies against the ␣4LG1-3 fragment demonstrated the expression and extracellular deposition of laminin ␣4 chain particularly in heart and skeletal muscle, in lung tissues, around fat and peripheral nerve cells, and in various vessel walls. They agree with previous expression data obtained by either in situ hybridization or by staining with antibodies generated against fusion proteins encoding domains II/I or the LG2-3 modules of the ␣4 chain (4,5,8,11). Surprisingly, no distinct staining could be obtained with antibodies against ␣4LG4 -5 in tissues that were otherwise strongly stained for ␣4LG1-3, ␣2LG1-3, and ␣2LG4 -5. This probably indicates loss of ␣4LG4 -5 from tissues after proteolytic release or, less likely, masking of its antigenic epitopes by interactions with other tissue components. Ultrastructural localizations by immunogold staining demonstrated the restriction of laminin ␣2 chains to basement membranes of skeletal muscle cells, pericytes, and endothelial cells. In heart, however, the ␣2 chains were not detected in the basement membrane around cardiomyocytes but instead in the interstitial microfibrillar matrix adjacent to them. The laminin ␣4 chain was not a constituent of endothelial or other basement membranes but was deposited in adjacent extracellular regions. Since endothelial cells produce ␣4 chain containing laminin, as shown here and previously (4), these proteins presumably diffuse away and contribute primarily to the connection of the outer regions of vessel walls to the extracellular matrix. It could also indicate the association of these laminins with microfibrils, which are known to contain fibulins (36,47,48).