Scaffold-forming and Adhesive Contributions of Synthetic Laminin-binding Proteins to Basement Membrane Assembly*

Laminins that possess three short arms contribute to basement membrane assembly by anchoring to cell surfaces, polymerizing, and binding to nidogen and collagen IV. Although laminins containing the α4 and α5 subunits are expressed in α2-deficient congenital muscular dystrophy, they may be ineffective substitutes because they bind weakly to cell surfaces and/or because they lack the third arm needed for polymerization. We asked whether linker proteins engineered to bind to deficient laminins that provide such missing activities would promote basement membrane assembly in a Schwann cell model. A chimeric fusion protein (αLNNd) that adds a short arm terminus to laminin through the nidogen binding locus was generated and compared with the dystrophy-ameliorating protein miniagrin (mAgrin) that binds to the laminin coiled-coil dystroglycan and sulfatides. αLNNd was found to mediate laminin binding to collagen IV, to bind to galactosyl sulfatide, and to selectively convert α-short arm deletion-mutant laminins LmΔαLN and LmΔαLN-L4b into polymerizing laminins. This protein enabled polymerization-deficient laminin but not an adhesion-deficient laminin lacking LG domains (LmΔLG) to assemble an extracellular matrix on Schwann cell surfaces. mAgrin, on the other hand, enabled LmΔLG to form an extracellular matrix on cell surfaces without increasing accumulation of non-polymerizing laminins. These gain-of-function studies reveal distinct polymerization and anchorage contributions to basement membrane assembly in which the three different LN domains mediate the former, and the LG domains provide primary anchorage with secondary contributions from the αLN domain. These findings may be relevant for an understanding of the pathogenesis and treatment of laminin deficiency states.

Basement membranes are specialized cell-adherent extracellular matrices consisting primarily of laminins, collagen IV, nidogens, and the heparan sulfate proteoglycans agrin and perlecan (for review, see Ref. 1). Among these, the laminins constitute a family of heterotrimeric glycoproteins that are essential for the assembly of basement membrane scaffolds (2,3). One property of laminin thought to be critical for basement membrane assembly is that of its anchorage to cell surfaces, a process that appears to be mediated through the LG domains of the ␣-subunit. Deletion of the five laminin-111 LG domains or of LG domains 4 -5 that contain dystroglycan and sulfatide binding loci or excess inhibiting LG4 -5 fragment was found to result in a failure of basement membrane assembly in an experimental Schwann cell model (4 -6). These studies further suggested that the reason laminin anchorage is crucial is that it provides the key linkage between the cell surface and the extracellular matrix scaffolding such that the other basement membrane components become tethered through laminin.
A second property of laminin is its polymerization into a network-like scaffolding (7,8). Laminin-111 (␣1␤1␥1), the most extensively studied in this regard, self-assembles in a thermally reversible manner with an initial oligomer-forming step followed by a calcium-dependent multimer-forming step (7). Laminin fragment and domain loss-of-function analyses have provided evidence that polymerization requires the participation of all three (␣, ␤, and ␥) LN domains located at the N termini of the short arms (6,9) such that laminins that possess fewer domains (as seen with truncated ␣3 and ␣4-laminins) lack the ability to polymerize (6,10).
A third property of laminin found to contribute to basement membrane assembly and stability is that of the binding of nidogen-1 and nidogen-2 (11)(12)(13). The nidogen-1 interaction is mediated between the laminin ␥1-LEb3 domain and the nidogen G3 domain. Nidogen G2 and G3 domains, in turn, bind to collagen IV. Although many basement membranes do not exhibit an absolute requirement of this bridging interaction, it appears likely that the interaction increases basement membrane stability (14 -16).
The principal laminins of Schwann cell endoneurial and skeletal muscle sarcolemmal basement membranes contain the ␣2-subunit (17). The absence of this subunit found in laminins 211 and 221 has been shown to cause a congenital muscular dystrophy and peripheral neuropathy in humans (classified as type MDC1A) and in mice (for review, see Ref. 18). Both defects have been corrected by transgenic expression of full-length laminin ␣1 subunit, indicating interchangeability of the ␣1 and ␣2 chains (19,20). A characteristic of ␣2 laminin-deficient congenital muscular dystrophy is a compensatory increase in the laminin ␣4 subunit both in nerve and muscle. The assembly and functions of ␣4-laminin in basement membrane are not well understood. The protein is thought to be non-polymerizing with low affinity binding for ␣-dystroglycan, sulfatides, and ␣6␤1 and ␣7␤1 integrins (21,22). Improved muscle function in laminin-deficient dystrophic mice, but not improved nerve function, was observed with transgenic expression of a internal domain-truncated muscle agrin (23,24) that binds to laminin and to ␣-dystroglycan (Denzer et al. 45 and Gesemann et al. 35). Although it is likely that the benefit of effect depends on these interactions, it is less clear whether amelioration of the muscle phenotype is due primarily to the enhancement of ␣4-laminin adhesion, to alterations of sarcolemmal ␣5-laminin, or to some other effect.
Cultured Schwann cells have provided a useful model with which to study basement membrane assembly (4 -6). Studies revealed that the galactosyl sulfatide present on the surface of these cells plays an important role in basement membrane anchorage through their binding to laminins, enabling laminindependent signaling through dystroglycan and ␤1-integrins (5,25). Furthermore, both laminin LN and LG domains were found to be required for laminin assembly of Schwann cell surfaces either in the absence or presence of nidogen and collagen IV (6). Interestingly, neither ␤1-integrins nor dystroglycan was required for laminin anchorage during initial basement membrane assembly on these cells (5). These receptors may instead act to link (and hence stabilize) the basement membrane to the underlying cell cytoskeleton (26).
In the current study we asked whether laminin deficits of polymerization resulting from LN domain deletions and/or deficits of cell surface binding resulting from LG-domain deletions could be corrected with laminin-binding proteins that add back missing domain activities. Such synthetic protein reagents could provide analytical tools to help understand the role of different domains in basement membrane assembly with the potential for the development of therapeutic approaches. Because there is a nidogen-binding site on laminin ␥1 chain near the intersection of the 3 short arms, we designed a chimeric protein (␣LNNd) containing the N-terminal ␣1 LN-LEa domains attached to nidogen-1 G2-rod-G3 domains. We evaluated the capacity of this synthetic short arm to bind laminin and provide type IV collagen binding in lieu of the nidogen it replaces and compared its behavior in a Schwann cell model of basement membrane assembly with that of muscle (non-neural) miniagrin. The fusion protein was found to specifically facilitate polymer formation and basement membrane accumulation of N-terminal-truncated ␣1-laminins on cultured Schwann cells, whereas miniagrin corrected adhesion deficits.
A stable cell line expressing ␣LNNd was supplemented with zeocin at 100 g/ml, whereas recombinant laminin lines were supplemented with puromycin, zeocin, and G418 at a final concentration of 1, 100, and 500 g/ml, respectively. Immunoprecipitation, SDS-PAGE, and Western blot analysis of secreted protein was used to confirm expression of trimeric laminin, ␣LNNd, nidogen-1, and mAgrin in the stable cell lines.
Co-purified laminin and ␣LN-Nd were isolated directly on FLAG M2-agarose and prepared as above. Type IV collagen and laminin-111 were extracted from lathyritic mouse EHS tumor and purified as described (31).
Protein Concentrations-EHS-laminin concentrations were determined by absorbance (280 nm) as described (9) with molarity determined based on the protein mass of 710 kDa. Absorbance was also used to measure the concentration of ␣LNNd and mAgrin with masses of 156 and 125 kDa, respectively. Protein mass and molar concentrations of laminins containing domain deletions (Lm⌬LG, 605 kDa; Lm⌬␣LN, 682 kDa; Lm⌬␣LN-L4b, 558 kDa; Lm⌬(␣LN&LG), 577 kDa) were determined by gel densitometry of their Coomassie Bluestained bands compared with those of EHS-laminin with corrections as needed for decreased mass.
Laminin Polymerization Assay-Aliquots (50 l) of laminin without or with ␣LNNd in polymerization buffer were incubated at 37°C, sedimented to separate polymerized protein, and analyzed by SDS-PAGE as previously described in detail (6). The apparent critical concentration was calculated from the product of x intercept and slope (10).
Assay of ␣LNNd Binding to Laminin and Collagen-IV-One g of recombinant laminin or collagen-IV was bound to a 96-well flat-bottomed dish (Nunc) in 40 mM sodium carbonate buffer overnight at 4°C. Plates were blocked with 1 mg/ml BSA in phosphate-buffered saline and 0.06%Triton X-100, then incubated with 2-fold increasing amounts of ␣LNNd or nidogen (0.013-16 g/ml). Protein was detected by entactinspecific (i.e. nidogen-1) monoclonal antibody (Chemicon MAB1946), protein A-horseradish peroxidase (Sigma P8651), and o-phenylenediamine (Sigma P3888). To determine collagen binding to ␣LNNdor nidogen-bound laminin, one g of recombinant laminin was coated onto a 96-well flat bottomed dish following by blocking with BSA and 5 g/ml nidogen or ␣LNNd incubation for 1 h at room temperature. After 3 washes with phosphate-buffered saline and 0.06%Triton X-100, increasing amounts of collagen-IV (0.01-10 g/ml) were added for 1 h at room temperature. Bound collagen was detected with collagen-IV-specific antibody (Chemicon AB 756P), protein A-horseradish peroxidase (Sigma P8651), and o-phenylenediamine (Sigma P3888) with absorbance measured at 492 nm with a TECAN Spectrafluor plate reader.
Sulfatide Binding Assay-HSO 4 -3Gal␤1-1ЈCeramide (brain galactosyl sulfatides) and galactosyl ceramide (Sigma C4905) were dissolved in methanol, and 0.1 g was added per immulon-1B microtiter well (ThermoLab systems). The plate was dried overnight at room temperature, and the wells were washed and blocked with enzyme-linked immunosorbent assay blocking buffer (1% BSA in TBS-50/Ca 2ϩ ). Proteins in varying concentrations in enzyme-linked immunosorbent assay blocking buffer were added to each well and incubated for 1 h at room temperature. Protein binding was detected with a horseradish peroxidase-linked monoclonal FLAG antibody (Sigma) and o-phenylenediamine (Sigma) with absorbance at 492 nm on a TECAN Spectrafluor.
Cell Culturing-Schwann cells isolated from sciatic nerves from newborn Sprague-Dawley rats were the kind gift of Dr. James Salzer (New York University). These cells were maintained in Dulbecco's modified Eagle's medium, 10% fetal calf serum (Gemini Bio Products), neuregulin (0.5 g/ml, Sigma), forskalin (0.2 g/ml, Sigma), 1% glutamine, and penicillinstreptomycin. Cells at passages 11-17 were plated onto 24-well dishes (Denville) and treated with the indicated proteins for 1 h at 37°C followed by washing and fixation. For electron microscopy (see the supplemental data) cells were plated in 60-mm Permanox dishes (Nalgene, Nunc) 2 days before the addition of proteins.

RESULTS
The domain and subunit composition of proteins used in this study are shown in Fig. 1. The proteins mAgrin, ␣LNNd, and Lm⌬(␣LN&LG) were characterized and compared with Lm⌬␣LN and wt Lm-111 by SDS-PAGE after purification.
Molecular Morphology-␣LNNd, mAgrin, Lm⌬␣LN-L4b, and wild-type (wt) laminin-111 alone and in complexes were visualized in electron micrographs after Pt/C rotary shadowing (Fig. 2). ␣LNNd molecules had the appearance of three globular domains separated by two short rods in either an extended or bent configuration. Lm⌬␣LN-L4b molecules had the appearance of a laminin with two rather than three short arms. After incubation of this laminin with ␣LNNd, a third short arm-like structure could be seen attached to the laminin from one of the two short arms near the junction of the other arms (Fig. 2, arrows), rendering laminin complexes not unlike (wt) laminin-111. mAgrin Pt/C replicas had the appearance of a flexible rod and globular protein in which four sphere-like structures could often be appreciated. After incubation of laminin-111 with mAgrin, the long arm (coiled-coil domain) was often seen to have a short projecting stub at about the mid-point along its length (arrows).
Binding Interactions of ␣LNNd-The chimeric protein was designed to possess three activities, i.e. binding of the C-terminal G3 domain to the laminin ␥1 domain LE3b, binding of domains G2 and G3 to type IV collagen, and the laminin ␣1-short arm component of polymerization that resides in the LN domain. Purified chimeric protein was evaluated for its ability to bind to laminin-111 and to type IV collagen (Fig. 3) in solid phase assays and found to bind to both. Chimeric ␣LNNd and nidogen-1 bound in an almost identical manner (apparent K D of 0.29 and 0.26 nM, respectively). ␣LNNd also bound to type IV collagen with an apparent K d of 4.0 nM (compared with 1.6 nM for nidogen-1). The ability of ␣LNNd was compared with that of nidogen-1 to mediate attachment of laminin to nidogen (ternary complex). ␣LNNd did this (K D of 1.4 nM) similar to nidogen-1 (1.4 nM).
The next question addressed was whether ␣LNNd, when bound to a laminin with two short arms, would mediate laminin polymerization (Fig. 4). This was evaluated in a standard selfassembly assay (37°C) in which the products are separated by sedimentation and evaluated by SDS-PAGE (6, 10). Lm⌬␣LN-L4b did not form a polymer when incubated alone. However, when Lm⌬␣LN-L4b and ␣LNNd were incubated in increasing equimolar concentrations, they co-sedimented in the polymer fraction in a concentration-dependent fashion with apparent critical concentrations (0.06 and 0.08 M) similar to that observed with laminin-111 (0.07 and 0.08 M). ␣LNNd did not appear to adversely affect the polymerization of (wt) laminin-  111 (critical concentration of 0.06 M) despite the addition of a fourth short arm creating a laminin complex with two ␣1LN domains. Incubation of Lm⌬␤LN or Lm⌬␥LN with ␣LNNd did not enable polymerization. This was interpreted as evidence that ␣LNNd is only able to rescue a polymerization deficit arising from deletion resulting in loss of the ␣LN domain (Lm⌬␣LN and Lm⌬␣LN-L4b).
Schwann cells contain galactosyl sulfatide that provides cell surface binding to laminins (5). To determine whether mAgrin and ␣LNNd bind to sulfatides, these proteins and different recombinant laminins were evaluated with a solid-phase assay (Fig. 5). Fitted half-maximal binding values (apparent K D ) were determined for mAgrin (0.06 M), ␣LNNd (0.02 M), (wt) lami-  , n ϭ 3). Half-maximal binding was fitted and found to be 1.4 nM for ␣LNNd and 1.5 nM for nidogen-1. Chimeric ␣LNNd, like nidogen-1, showed binding activity for both laminin and type IV collagen and was able to mediate formation of ternary complexes. . These changes are thought to primarily reflect the combined contributions arising from adding a second sulfatide binding domain to the mutant laminins possessing one binding domain, rather than polymerization, as detection of contributions of the latter is thought to require mobile lipid molecules reconstituted in a mobile bilayer (37). Binding (half-maximal and maximal) for laminin and (especially) mAgrin were found to be decreased when sulfated-galactosyl ceramide was diluted into non-sulfated galactosyl ceramide. Structural analysis of the laminin ␣1 LG4 domain has revealed several lysines and arginines that interact with the small sulfatides and is consistent with the hypothesis that each LG domain engages several glycolipid sulfates (38). Therefore, the decreases in binding in the assay may arise from the loss of sulfate charge density available to bind to the LG4 protein patch. Although the sulfatide composition of Schwann cell surfaces has not been determined, it is likely that it much less than 100%, and therefore, the effective laminin and agrin affinities may be lower than that detected with pure lipid. A caveat is that if the sulfatides are organized into compact rafts, the higher affinities could be preserved.
Laminin Accumulation on the Surface of Cultured Schwann Cells-It was previously found that Schwann cells treated with exogenous laminins assemble a thin basement membrane on the free cell surface in a process dependent upon interactions of the LG domains with cell surface sulfatides and upon the ability of laminins to polymerize (5,6). Laminin assembly was found in turn to enable the incorporation of collagen IV in the presence of nidogen-1. To determine whether ␣LNNd and mAgrin were capable of affecting laminin assembly on cells through their respective capacities to alter polymerization and adhesive interactions, Schwann cells were incubated with laminins bearing deletions of different domains in either the presence or absence of the above laminin-binding proteins (Figs. 6 -8).
Cells near confluency were incubated for an hour with proteins diluted into the culture medium, washed, fixed, and treated with antibody to laminin followed by detection with a fluorescent secondary reagent. The average cell fluorescent intensities (average and S.D. of sums of pixel intensities/cell) were determined from the antibody-stained images and plotted (Figs. 7 and 8). Treatment of cells with laminin-111 (wt) resulted in the accumulation of laminin (Figs. 6, A and B, and 7A) as reported previously (6). In contrast, treatment with the non-polymerizing laminins Lm⌬␣LN-L4b, Lm⌬␣LN, or the chimeric fusion protein ␣LNNd (detected with nidogen-specific antibody) resulted in almost no detectable protein (Fig. 6, E and I). However, when Lm⌬␣LN-L4b or Lm⌬␣LN was mixed with ␣LNNd in equimolar concentrations (14 nM), laminin fluorescence was substantially increased (Fig. 6, F and J, and Fig. 7, A-C) to levels approaching (ϳ60 -70% in different experiments) of (wt) laminin. Laminin accumulation on cells increased as a function of increasing concentration (Fig. 7A) and was accompanied by a corresponding increase in nidogen epitope (B) located on ␣LNNd. ␣LNNd was specific in its ability to improve laminin accumulation on cells as it had no appreciable effect when incubated with laminins bearing deletions of the ␤-LN or ␥-LN domains or bearing an incomplete complement of ␣-, ␤-, and ␥-LN domains or an incomplete set (␣-␣-␤ and ␣-␤-␤) of domains when combined with ␣LNNd (Fig. 7C). ␣LNNd caused a slight increase in wt laminin fluorescence that could be related to the small increase seen in the polymerization slope (Fig. 7D). The contribution of mAgrin to the accumulation of laminins on Schwann cells was also examined (Figs. 6, C, D, G, and H, and  8). Lm⌬LG failed to accumulate on cells, even at concentrations as high as 28 nM (Fig. 8A). Similarly, very little mAgrin was detected on cell surfaces when added without laminin (Fig. 6G). In contrast, coincubation of Lm⌬LG with mAgrin resulted in the accumulation of both laminin and mAgrin (Fig. 6, D and  H) in a concentration-dependent fashion (Fig. 8, A and B). The addition of mAgrin to (wt) laminin caused only a slight increase over that seen with the laminin alone (Fig. 8C). On the other hand, the addition of mAgrin to a non-polymerizing laminin did not lead to increased laminin accumulation (Fig. 8D).
The ultrastructure of cells treated with wt laminin, Lm⌬␣LN-L4b, Lm⌬LG, Lm⌬␣LN-L4b ϩ ␣LNNd, and Lm⌬LG ϩ mAgrin was evaluated after incubation of the components (14 nM) for one hour (supplemental Fig. S1). Thin basement membrane-like linear electron-dense matrices (lamina densa), separated from the cell surface by an electron lucid zone (lamina lucida), were detected in lengths of several m or more after treatment with ␣LNNd ϩ Lm⌬␣LN-L4b or with mA ϩ Lm⌬LG that was similar to the matrix formed with (wt) laminin and that was absent after treatment with the defective laminins in the absence of the synthetic linker proteins.
Because there appear to be separate polymerization and adhesive contributions required for laminin assembly on Schwann cell surfaces, it seemed reasonable to expect that a laminin that lacked both an LN domain and LG domains would be able to accumulate on cells only in the presence of both ␣LNNd and mAgrin. The recombinant laminin Lm⌬(␣LN&LG) was generated to evaluate this possibility. When Lm⌬(␣LN&LG), copurified with ␣LNNd, was incubated with cells, the laminin failed to accumulate at different concentrations (Figs. 6K and 8D). However, when ␣LNNd-Lm⌬(␣LN&LG) was also mixed with mAgrin, laminin accumulation was observed on cells (Fig. 6L) in a concentrationdependent fashion (Fig. 8, D  and E).
We asked whether laminins lacking LG domains can accumulate on cells by co-polymerization with intact laminins. To address this possibility, Lm⌬LG (20 g/ml, Myc-tagged) was mixed with (wt) laminin (20 g/ml and compared with Myc-tagged (wt) laminin (40 g/ml). The laminins were co-stained with antibodies to Myc and to LG4 -5 (anti-E3). Lm⌬LG was not detected in the mixture, whereas both wt laminins were detected (Fig. 6, O and P). Furthermore, a colinear increase of total and wt laminin was observed when wt laminin was varied with Lm⌬LG with the sum of the two laminins maintained constant (Fig. 8F). These data indicate that under the conditions of the assay (below the free solution critical concentration), all laminins are attached to the cell surface through LG domains. It follows that the cell surface extracellular matrix contains a laminin monolayer and that every Lm⌬LG that accumulates in the presence of mA must be simultaneously bound to mA and anchored to the cell surface.
detected on cell surfaces after treatment with collagen IV in the presence of either wt laminin with nidogen-1 or nonpolymerizing laminin with ␣LNNd. The collagen levels detected with the latter were substantial ( 1 ⁄2 to 3 ⁄4 of wt laminin plus nidogen). When mAgrin was incubated with Lm⌬LG, nidogen, and collagen, near-normal (i.e. wt) levels were detected (Fig. 9C). In agreement with earlier observations (6), non-polymerizing laminin, itself retained at very low levels relative to wt laminin, can maintain substantial levels of collagen so long as the nidogen-bridging interaction is preserved. Here ␣LNNd was able to substitute for nidogen for this activity.
The chimeric fusion protein ␣LNNd uses the nidogenbinding site to attach to laminin and, therefore, would be expected to be in competition with any existing nidogen. The effect of such competition was examined in the absence and presence of collagen IV (Supplemental Fig. S2). When ␣LNNd was varied to molar excess in the presence of 13 nM nidogen-1 and Lm⌬␣LN-L4b, laminin accumulation on cells increased with a plateau reached by ϳ25 nM. When nidogen-1 was varied to molar excess over constant 13 nM ␣LNNd and Lm⌬␣LN-L4b, laminin accumulation decreased with a low plateau reached above ϳ25 nM. However, the addition of 9 nM collagen IV dampened this effect such that the addition of molar excess of nidogen caused no appreciable decrease in laminin accumulation. An interpretation compatible with the results is that collagen IV forms a network that can accommodate substantial excess of nidogen or ␣LNNd, resulting in little apparent competition.

DISCUSSION
A study of two synthetic linker proteins that add functional activities to laminins and restore the ability of deficient laminins to assemble a basement membrane-like extracellular matrix on Schwann cells has provided insights into the mechanisms of basement membrane assembly. A model (Fig. 10) consistent with these findings and supported by our earlier studies is that a laminin initiates assembly by attaching to the cell surface through sulfatides (and also to ␣-dystroglycan to the degree to which it is present) and by forming linkages among adjacent laminins through polymerization. Nidogen-1 binds to laminin and also to type IV collagen, increasing the surface concentration of the latter and thereby promoting its own polymerization. Agrins further stabilize the laminins by binding to them and to the cell surface.
The study adds several elements to our understanding of laminin interactions and assembly. First, the data support, through a gain-of-function analysis, that the laminin poly-mer is formed by the binding of ␣, ,␤ and ␥ LN domains into a ternary domain complex. An absent ␣LN domain, but not an absent ␤LN or ␥LN, could be replaced with the missing ␣LN domain with restoration of self-assembly. It is interesting that placement of the synthetic linker arm at the nidogen-binding site located in domain LEb3 of the ␥-subunit near the intersection of the three short arms created a third arm sufficiently similar to the defective native arm to provide the activity. Furthermore, the internal L4a, LEb, L4c, and LEc of the ␣-subunit are largely dispensable for polymerization. These internal domains may serve primarily to add length to the short arm, affecting the spacing of laminins within the polymer but might also contribute to polymer stability. Of note, binding of ␣LNNd to wt laminin, which adds a fourth short arm duplicating the ␣LN domain, was not deleterious for polymerization and slightly enhanced it. The polymerization slope increase seen with recombinant and to a lesser extent with EHS-laminin may be the consequence of a fractional reduction of activity within the LN domain. , n ϭ 7, each condition). Laminin accumulation in the presence of collagen IV required either a polymerizing laminin or non-polymerizing laminin ϩ ␣LNNd. Collagen IV accumulation required only a non-polymerizing laminin ϩ nidogen or nonpolymerizing laminin ϩ ␣LNNd. Panel B, plots of mAgrin and Lm⌬LG contributions. Type IV collagen accumulation was increased with Lm⌬LG ϩ mAgrin ϩ nidogen (n ϭ 7). Panel C, mAgrin accumulation required Lm⌬LG in the presence of nidogen and collagen IV (n ϭ 5). mA, mAgrin. A study of interactions between laminin LN-LEa fragment pairs, in which binding was detected for ␣1LNLEa-␣1LNNEa pairs, led to a less-restricting hypothesis of assembly in which a laminin polymer could form with ternary complexes that lacked an ␣-␤-␥ composition (39). However, a subsequent domain loss-of-function analysis conducted with heterotrimeric laminins failed to support this alternative (6). The current study now provides gain-of-function evidence that a strict ␣-␤-␥ short arm complex is required. The possibility that the ␣-␣ subunit interaction is involved in the attachment between laminin polymer layers rather than polymerization per se presents itself as an interesting alternative hypothesis to explain its self-binding.
The linker protein ␣LNNd was able to largely, but not completely, restore laminin assembly on cell surfaces when coupled to laminin molecules lacking the ␣LN domain or entire ␣-short arm. The incompleteness of the rescue may be a consequence of instability in the recombinant linker protein, incorrectness of the length of the linker placement of ␣LN to the other LN domains, or a missing contribution from the internal ␣-short arm domains. Although the lack of change of the critical concentration for the linked laminin compared with wt laminin is more compatible with the first interpretation, further study will be required to resolve the question. Preservation of the type IV collagen binding sites of nidogen G2 and G3 domains within ␣LNNd allowed for the linked laminin complex to recruit type IV collagen to the cell surface in the absence of nidogen. The ability of nidogen to compete for ␣LNNd accumulation on cells was considerably reduced in the presence of type IV collagen. mAgrin, an internally truncated protein that binds strongly to the coiled-coil domain of laminin, to sulfatides, and to ␣-dystroglycan can also enhance basement membrane assembly. In particular, we found that it enables the poorly adhesive laminin Lm⌬LG to become anchored to the cell surface, assemble, and recruit nidogen and type IV collagen. Anchorage, however, was found to be sufficient for such assembly on Schwann cells only in the presence of LN-mediated polymerization. This limitation was revealed by the specificities of laminin rescue and with a laminin that lacked both an ␣-LN domain and LG domains with assembly restoration only if mAgrin and ␣LNNd were coincubated with the doubly truncated Lm⌬(␣LN&LG).
Earlier analysis of the cultured Schwann cells revealed that sulfatides constitute the principal contribution for laminin anchorage (to be distinguished from signaling contributions) and not ␣-dystroglycan, ␤1-integrins, or heparan sulfates 5). In another (breast epithelial) cell line, a greater dystroglycan contribution for laminin accumulation was reported (25). The explanation for the observed differences may lie in the relative surface density of these different molecules, all capable of laminin binding through LG domains, i.e. in Schwann cells there are too few dystroglycan molecules available for laminin binding compared with available sulfatide molecules for dystroglycan to serve as principal anchor. However, it seems a reasonable expectation that in some tissues dystroglycan (notably in muscle) or integrins will be found to serve as the chief anchoring species because of their abundance. The significance of these receptors is that they can link the basement membrane to the underlying cortical cytoskeleton to stabilize the basement membrane-cell interface. mAgrin is an interesting linker protein in that it binds to ␥1-laminins, ␣-dystroglycan, sulfatides, and an integrin and may well be suited to provide an anchorage function for laminins in different cellular environments.
A model refinement to be considered is based on the observations that sulfatide binding contributions also arise from ␣LN domains (this study and Garbe et al. 36). These domains may provide supplemental adhesion of laminin molecules to the cell surface such that contacts form with both LG and ␣LN. Nonetheless, adhesion through LG must occur for laminin to accumulate on cell surfaces. Furthermore, polymerization is required in addition to ␣LN and LG for significant assembly on cell surfaces and cannot be accomplished with laminins that only possess ␣LN and LG domains.
Both this study and that of McKee et al. (6) have revealed a contribution of type IV collagen and nidogen in which a collagen-rich laminin-poor discontinuous extracellular matrix can form on a cell surface in the absence of laminin polymerization and presence of collagen and nidogen. The collagen levels achieved under these circumstances have been found to be about half that with a polymerizing laminin. In contrast, cell surface of laminin or collagen has not been observed to any appreciable degree with a non-adhesive laminin (without LG domains) on Schwann cells.
In summary, domain-modified laminin-111 proteins were used in this study to identify key self-assembly and anchoring activities that distinguish the two synthetic linker proteins and to analyze their contributions to basement membrane assembly in a model culture system. A question that arises is how predictive are these findings for complex basement membranes of different tissues? Of particular interest, from a human disease standpoint, are the basement membranes of the Schwann cell endoneurium in nerve and the sarcolemma in skeletal muscle. Both of these basement membranes are defective in the MDC1A congenital muscular dystrophies and mouse models that result from null, hypomorphic, and domain-altering mutations of the gene coding for the laminin ␣2 subunit. Laminin-211 is the principal laminin of these basement membranes. However, ␣4and ␣5-laminins are expressed in the dystrophy (23, 40 -42). ␣4-laminins lack the ␣-subunit short arm for polymerization and bind less well to ␣-dystroglycan and sulfatides (21,22), whereas laminin-511, thought to polymerize, has also reduced binding to these components (43). What then is the basis for a mAgrin rescue of muscle and its failure (so far) to rescue in nerve (23,24,40)? One possible explanation for the observation in peripheral nerve is that mAgrin cannot rescue an ␣2-defect through ␣4-laminin. The phenotypic rescue in muscle, on the other hand, is substantial. It has been suggested that mAgrin accomplishes this in muscle through its interaction with ␣4-laminins and dystroglycan (23). This would appear to be a reasonable interpretation if the only defect is in adhesion. However, if laminin polymerization is also important, then the rescue of ␣4-laminin would be insufficient. An alternative possibility, consistent with the findings of this study, is that the rescue is mediated through ␣5-laminins, a laminin found to increase in the sarcolemmal basement membrane after mAgin treatment (23). In nerve, there may be too little ␣5-laminin available for binding to mAgrin to mediate rescue of radial sort-ing. A caveat to consider in comparing the effects of mAgrin on nerve and muscle is that these tissues may contain different densities of anchors/receptors that could differentially effect a requirement for polymerization. The model does not rule out the possibility that a sufficiently high density distribution of a high affinity surface component could reduce the requirement for polymerization by enabling binding of a sufficiently dense distribution of laminin molecules to form a stable matrix in the absence of the contribution provided by polymerization (6).
A related question is whether ␣LNNd, like mAgrin, holds potential to ameliorate the radial sorting defect and muscle pathology seen in laminin-␣2 deficiency states. The most obvious situation in which one might expect ␣LNNd to beneficially affect nerve and muscle is with the dy 2J dystrophic mouse that arises from an in-frame deletion within the ␣2LN domain (44). In those variants of the syndrome in which there is little or no laminin ␣2 expression, one might predict that such amelioration would be less likely, especially if the repair is mediated by binding of ␣LNNd to laminin-411. The deficit of anchorage for this laminin would remain uncorrected. On the other hand, the combined action of mAgrin and ␣LNNd might efficiently convert laminin-411 into a strongly adhesive polymerizing laminin.