Alternative splice variants of alpha 7 beta 1 integrin selectively recognize different laminin isoforms.

The integrin alpha(7)beta(1) occurs in several cytoplasmic (alpha(7A), alpha(7B)) and extracellular splice variants (alpha(7X1), alpha(7X2)), which are differentially expressed during development of skeletal and heart muscle. The extracellular variants result from the alternative splicing of exons X1 and X2, corresponding to a segment within the putative ligand binding domain. To study the specificity and affinity of the X1/X2 variants to different laminin isoforms, soluble alpha(7)beta(1) complexes were prepared by recombinant coexpression of the extracellular domains of the alpha- and beta-subunits. The binding of these complexes to purified ligands was measured by solid phase binding assays. Surprisingly, the alternative splice variants revealed different and specific affinities to different laminin isoforms. While the alpha(7X2) variant bound much more strongly to laminin-1 than the alpha(7X1) variant, the latter showed a high affinity binding to laminins-8 and -10/11. Laminin-2, the major laminin isoform in skeletal muscle, was recognized by both variants, whereas none of the two variants were able to interact with laminin-5. A specific blocking antibody inhibited the binding of both variants to all laminins tested, indicating the involvement of common epitopes in alpha(7X1)beta(1) and alpha(7X2)beta(1). Because laminin-8 and -10/11 as well as alpha(7X1) are expressed in developing skeletal and cardiac muscle, these findings suggest that alpha(7X1)beta(1) may represent a physiological receptor with novel specificities for laminin-8 and -10.

The laminin-binding integrin ␣ 7 ␤ 1 is the major laminin receptor of skeletal, cardiac, and smooth muscle cells (for review, see Ref. 1), but it has also been detected in some human melanoma cells (2) and glioblastoma cells (3). In skeletal muscle, it is located predominantly in myotendinous and in neuromuscular junctions (4,5) and in the sarcolemma (6). Experimental evidence from several laboratories indicates that ␣ 7 ␤ 1 integrin is involved in laminin-induced migration of skeletal myoblasts (7,8) and other cells: cell motility of melanoma, HEK293 cells (9,10), and MCF7 cells (11) on laminin-1 and -2 was considerably enhanced after transfection with ␣ 7 integrin expression vectors. The critical role of ␣ 7 integrins in muscle function became evident after inactivation of both alleles of the ␣ 7 gene itga7 in mice. Deficient mice developed a novel form of myopathy, accompanied by disruption of the myotendinous junctions (12). Subsequently, human patients were identified with similar defects associated with mutations in the Itga7 gene (13).
The ␣ 7 integrin subunits are expressed in two cytoplasmic and two extracellular splice variants in human and mouse tissues (14 -16), the expression of which is developmentally regulated. The cytoplasmic variant ␣ 7B is expressed in cardiac and smooth muscle, skeletal myoblasts, and embryonic and adult skeletal muscle, whereas the ␣ 7A variant is expressed only in mature skeletal muscles but not in cardiac muscle (6). The extracellular variants ␣ 7X1 and ␣ 7X2 result from alternative splicing of exons coding for a region between the homology repeats III and IV, located in the putative ␤-propeller domain of ␣ 7 (17), and represent part of the putative ligand binding site. The X1 and X2 splice variants are expressed in equal amounts in mouse skeletal myoblasts and adult heart, whereas in adult skeletal muscle, mainly the ␣ 7X2 splice variant can be found (14,15). The specific roles of these splice variants are still unclear. After experimental clustering of acetylcholine receptors on the surface of ␣ 7X1 -or ␣ 7X2 -transfected C2C12 myotubes by laminin-1, only the ␣ 7X2 was detected within the acetylcholine receptor clusters (18). Both splice variants bind to laminin-1-and laminin-2-coated surfaces when expressed in HEK293 cells; however, ␣ 7X1 B promoted cell migration only on laminin-2, and ␣ 7X2B stimulated motility on both substrates (10). Further, the cell-specific environment seems to have a critical impact on binding (10,17).
In order to define the affinity and specificity of ␣ 7 ␤ 1 integrin and its extracellular splice variants for different laminins, soluble ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 complexes were expressed recombinantly in HEK293 cells, and affinities for laminin isoforms were analyzed using a solid phase binding assay. The extracellular domains of integrin ␤ 1 and ␣ 7 formed heterodimers and were secreted to the culture medium. Surprisingly, X1 and X2 variants showed distinct binding affinities for different laminin isoforms. The X2 splice variant showed a high affinity binding to laminin-1 and its E8 fragment but did not bind laminins-8 and -10/11. The X1 variant, in contrast, specifically recognized laminin-8 and -10, and to a lesser extent, it also recognized laminin-1. Both X1 and X2 showed similar affinities for laminin-2, whereas laminin-5 was not recognized by either splice variant. Because striking similarities exist between the expression patterns of the ␣ 7X1 ␤ 1 splice variant and laminins-8 and -10, we assume that ␣ 7X1 ␤ 1 represents a potential receptor for laminin-8 and -10 during early steps in embryogenesis.

EXPERIMENTAL PROCEDURES
Purification of Laminins and Other Matrix Proteins-Laminin-1 was purified from the murine Engelbreth-Holm-Swarm tumor (30), and laminin-2 was extracted from mouse heart muscle and purified as described previously (31). Laminin-5 was prepared from human SSC25 cells (32) (kindly provided by Dr. J. Eble, University of Mü nster, Germany). Bovine laminin-8 was isolated from conditioned medium of bovine aortic endothelial cells endothelial cells as described in Ref. 23. Laminin-10/11 was purified from human placenta by affinity chromatography using the anti-␣ 5 antibody, 4C7 (33). Collagen I and fibronectin were purchased from Sigma.
Generation of Expression Vectors for Soluble Integrin Chains-The cloning of the murine ␣ 7BX2 variant was described earlier (9) with the full-length cDNA corresponding to positions 14 -3829 of GenBank TM accession number GB:NM_008398. An additional RGS-His 6 tag (Qiagen) was introduced at the C-terminal end of the extracellular domain (position 3280) using a PCR-based strategy, resulting in the construct pUC18/␣ 7X2 -E-His. The extracellular domain of the X1 variant of ␣ 7B was generated by a reverse transcription-PCR fusion strategy using overlapping primers corresponding to the X1-specific sequence (14,15) in combination with N-and C-terminal primers. The resulting 1060-bp N-terminal fragment was cloned and fused with the C-terminal half containing the RGS-His 6 resulting in pUC18/␣ 7X1 -E-His. The plasmid pBSmint␤ 1 -N contains the full-length cDNA of the murine ␤ 1A cDNA (a kind gift of S. Johansson and R. Faessler), and sequences coding for the C terminus of the ectodomain were amplified by PCR using Vent Polymerase (Biolabs) with an additional Strep-tag sequence, SAWRH-PQFGG (34), introduced at the C terminus at position 2278. The 0.44-kb fragment, corresponding to positions 1880 -2278 (GenBank TM accession number GB:mmintbr), was cloned and fused via an internal BglII site with the N-terminal HhaI-BglII fragment (positions 57-1910) of pBSmint␤ 1 -N, resulting in pBS/␤ 1 -E-Strep. All clones were verified by restriction mapping and sequencing. The inserts from pUC18/␣ 7X1 -E-His and pUC18/␣ 7X2 -E-His, coding for His-tagged ␣ 7 ectodomains, were isolated by XbaI/HpaI digestion and recloned in the corresponding sites of the vector pCMVsis (35), resulting in pCMV/␣ 7X1 -E-His and pCMV/␣ 7X1 -E-His. The insert from pBS/␤ 1 -E-Strep was recloned in the expression vector pcDNA3 (Invitrogen) by the use of KpnI/XbaI, resulting in pcDNA/␤ 1 -E-Strep.
Purification of Soluble ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 Heterodimers-Transfected HEK293 cells were grown to confluency and incubated for a further 2 days in serum-free medium. Medium was harvested, 0.04% sodium azide was added, and the samples were concentrated by ultrafiltration (Amicon Inc.). Nickel-nitrilotriacetic acid agarose, equilibrated with 50 mM sodium phosphate, 300 mM NaCl, pH 7.8, was added, and the suspension was incubated for 1-16 h at 4°C. After centrifugation, the nickel-nitrilotriacetic acid agarose was washed with equilibration buffer containing 5 mM imidazole and finally eluted with 50 mM imidazole in the same buffer. The eluted material was concentrated in a Centricon 30 ultrafiltration unit (Amicon Inc.), washed three times with Tris-buffered saline (TBS), 1 and adjusted to 1 mM MnCl 2 and 2 mM MgCl 2 (TBS/Mn/Mg). Up to 1.5 mg/liter of purified complexes could be isolated from conditioned media.
Solid Phase Binding Assays-Extracellular matrix proteins in TBS were coated onto 96-well microtiter plates (Maxisorb, Nunc) at 4°C overnight. The efficiency of coating with different laminin variants was tested by the binding of 125 I-labeled proteins and showed comparable and concentration-dependent binding of all laminin isoforms to the plastic surface (33). All the following steps were performed at room temperature. After coating, plates were washed with TBS/Mn/Mg, and nonspecific binding sites were blocked with 2% bovine serum albumin in TBS/Mn/Mg for 1 h. Soluble ␣ 7 ␤ 1 integrin was diluted with 1% bovine serum albumin in TBS/Mn/Mg, added to the coated plates, and incubated for 2 h. The plates were washed three times with TBS/Mn/Mg, and bound receptor was detected using either mouse monoclonal antimouse ␣ 7 antibody, 3C12 (10), or a rabbit polyclonal anti-rat ␤ 1 antibody (36). After incubation for 1 h and washing with TBS/Mn/Mg, secondary antibody (anti-mouse or anti-rabbit) coupled to horseradish peroxidase was applied and incubated for 1 h. ABTS substrate (Roche Molecular Biochemicals) was used for detection, and absorbance was measured at 405 nm. All measurements were performed in duplicates or triplicates. Mean value and standard deviation are shown for one out of three-five typical experiments. Calculations of K d values were done by methods based on Scatchard plot or saturation plot analysis according to Ref. 37. For quantitation of the ␣ 7 ␤ 1 ectodomains, calibration curves were generated using a sandwich ELISA. Microtiter plates were coated with 50 g/ml of the monoclonal anti-␣ 7 antibody 3C12 (10), and serial dilutions of purified recombinant ectodomains were added. Bound ␣ 7 ␤ 1 complexes were detected by the same anti-␤ 1 antibody used during solid phase assays. The calibration curves were linear up to 0.1 g of isolated complexes. K d values were calculated according to Ref. 37 from experiments where saturation of binding could be achieved.

RESULTS
The Extracellular Domains of ␣ 7 and ␤ 1 Integrin Subunits Form Stable Heterodimers-To study the affinity and specificity of the ␣ 7 integrin splice variants, X1 and X2, the extracellular domains were coexpressed together with ␤ 1 integrin after deletion of the transmembrane and cytoplasmatic domains. A (His) 6tag coding sequence was fused to the C termini of the truncated ␣ 7X1 and ␣ 7X2 variants to facilitate their purification. After cotransfection into HEK293 cells, stable clones that secreted soluble ␣ 7X1 ␤ 1 or ␣ 7X2 ␤ 1 integrin heterodimers were identified by immunoblotting of supernatants. Soluble ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 complexes were purified from the cell culture medium using nickelnitrilotriacetic acid affinity chromatography (Fig. 1). The chain composition of the purified complexes was tested by immunoblotting using specific antibodies to ␣ 7 and ␤ 1 subunits (Fig. 1, B and  C). The estimated sizes for the soluble ␣ 7X1 (97 kDa) and ␣ 7X2 (95 kDa) subunits corresponded with the calculated sizes of the expressed domains, whereas the diffuse band obtained with the ␤ 1 subunit (110 -120 kDa) is indicative for a high level of glycosylation. Although most clones produced and secreted ␤ 1 chains in excess, both truncated subunit chains appeared in comparable amounts after purification, although the presence of small amounts of single chains cannot be excluded (Fig. 1A). Mn 2ϩ was added to purified complexes and was always present during experiments. The absence of Mn 2ϩ resulted in a loss of stability of the ␣ 7 ␤ 1 complexes; thus Mn 2ϩ could not be replaced by other divalent anions. Up to 1.5 mg of soluble complexes were isolated from 1 liter of conditioned medium.
␣ 7X2 ␤ 1 Binds to Laminin-1 with Higher Affinity Than ␣ 7X1 ␤ 1 -The binding of the ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 variants to laminin-1 was investigated by a solid phase binding assay on microtiter dishes coated with purified laminin-1 at varying concentrations (2.5-20 g/ml). Bound receptor was detected by antibodies specific for the ␣ 7 (not shown) or the ␤ 1 chain ( Fig.  2A). The binding was saturable and revealed a higher affinity of ␣7 X2 ␤ 1 for laminin-1 as compared with ␣ 7X1 ␤ 1 . The difference was further substantiated by calculations of the K d values using Scatchard plot and saturation plot analysis (37). Amounts of bound ectodomain complexes were calculated from calibration curves established by a sandwich ELISA (see "Experimental Procedures"). The affinity of ␣ 7X1 ␤ 1 for laminin-1 (K d : ϳ17 ϫ 10 Ϫ9 M) is significantly lower than the affinity of the ␣ 7X2 ␤ 1 variant for the same ligand (K d : ϳ3 ϫ 10 Ϫ9 M). Corresponding differences were also seen after the binding of variable concentrations of soluble receptors of ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 to constant concentrations (10 g/ml) of laminin-1 (Fig. 2B). Similarly, ␣ 7X2 ␤ 1 showed a significantly higher affinity for the E8 fragment of laminin-1 than the X1 splice variant (Fig. 3). In contrast, both complexes bound to laminin-2 to a comparable extent at high concentrations (100 nM) of receptors (Figs. 2A and 3), which is in agreement with a previous study where we found comparable levels of binding of ␣ 7X1 -and ␣ 7X2 -expressing HEK293 cells to laminin-2 (10). Low concentrations of receptor revealed a higher affinity of ␣ 7X1 ␤ 1 for LN-2 (Fig. 2B). Neither ␣ 7X1 ␤ 1 nor ␣ 7X2 ␤ 1 bound to laminin-5 (Fig. 3), which is recognized by ␣ 3 ␤ 1 and ␣ 6 ␤ 4 (32,38,39). No binding of either ␣ 7 splice variant to collagen I, fibronectin, or bovine serum albumin could be detected. The presence of EDTA completely inhibited the binding of receptor variants to all laminins (data not shown).
The ␣ 7 -specific antibody 6A11 was previously shown to inhibit ␣ 7 ␤ 1 -mediated adhesion of transfected HEK293 cells (10). The addition of this blocking antibody inhibited the binding of ␣ 7X2 to laminin-1 as well as ␣ 7X1 to laminins-1, -8, and -10/11 (Fig. 4), indicating that common interaction sites are involved in ligand binding of both splice variants. DISCUSSION In this study, we present for the first time evidences for differential specificity of the extracellular splice variants of ␣ 7 integrin for laminin isoform ligands. Laminins as biologically active components of basement membranes have been shown to regulate cell adhesion and spreading, cell proliferation, differentiation, and migration. Many in vitro studies on the biologi- and bovine serum albumin (BSA) were coated, and the binding of soluble receptors was determined by ELISA. The coating concentration was 10 g/ml, except for Ln-1/E8 (3 g/ml).

FIG. 4.
Interaction of soluble ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 complexes to laminins can be blocked by the ␣ 7 -specific antibody 6A11. Laminin-1,-8, and -10/11 were coated at 5 g/ml and incubated with 50 nM ␣ 7X1 ␤ 1 either in the absence (white) or presence (black) of the inhibitory antibody 6A11 (A). Laminin-1 was coated at 2-10 g/ml to plates and incubated with 5 or 50 nM soluble ␣ 7X2 ␤ 1 complexes either in the absence (white) or presence (black) of 10 g/ml of antibody 6A11 (B). cal roles of laminin have been performed using laminin-1, predominantly found in embryonic and fetal tissues but absent from many adult tissues including skeletal muscle (20 -22). At present, 15 distinct laminins have been described in different basement membranes (26,29,40,41). The major laminin isoform of skeletal muscle is laminin-2, whereas laminin-4 is restricted to neuromuscular junctions (25). Basement membranes surrounding embryonic myofibers also contain laminins-8 and -10, which are down-regulated at this site with development.
The ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 integrin complexes used in this study were prepared as recombinant, soluble ectodomains lacking the cytoplasmic and transmembrane domains. Despite the absence of these domains, the extracellular domains of ␣ 7X1 or ␣ 7X2 subunits formed stable heterodimeric complexes with the truncated ␤ 1 subunit, which apparently retained their native structure and ligand binding capacity. Heterodimerization of ␣ 7 and ␤ 1 subunits was proven by immunoprecipitation of the complexes with antibodies to ␣ 7 , as well as by affinity chromatography of the tagged ␣ 7X1 or ␣ 7X2 subunits. Obviously, the extracellular domains of ␣ 7X1 /␣ X2 and ␤ 1 contain sufficient structural information for their assembly into stable heterodimers. Alternatively, soluble ␣ 3 ␤ 1 heterodimers devoid of cytoplasmic and transmembrane domains have been prepared by coupling jun and fos leucine zippers to the C termini of subunits to stabilize the complex (32). In that case, the use of dimerization domains stabilized the complex and also allowed binding studies in the presence of different cations. The essential addition of Mn 2ϩ for the stabilization of the soluble ␣ 7 ␤ 1 complexes points into this direction. The method used in this study may be useful for the detailed analysis of receptor-ligand interactions by mutagenesis studies and epitope mapping.
The binding of ␣ 7X1 ␤ 1 to laminin-8 and -10/11, as well as the interaction of both variants to laminin-1 and -2, could be completely inhibited by the ␣ 7 blocking antibody, 6A11, which was shown previously to block ␣ 7X2 -mediated binding of cells to laminin-1/E8 fragment (10). This indicates that the ligand binding of ␣ 7X1 /␣ 7X2 variants with laminin isoforms share common epitopes in the extracellular domain, whereas specificity is defined by the X1/X2 region located between homology repeats III and IV in the ␤-propeller of the ␣ 7 chain (44). Nevertheless, it cannot be excluded that the blocking effects of the antibody may result from steric hindrance or by allosteric effects.
The soluble ␣ 7X2 ␤ 1 showed a higher affinity for laminin-1 and laminin-1/E8 fragment than the ␣ 7X1 ␤ 1 complex. In accordance with this finding, ␣ 7BX2 -transfected HEK293 cells are characterized by an increased motility on laminin-1 as compared with ␣ 7BX1 -transfected cells, although adhesion rates were about equal for both splice variants (10). After prolonged attachment times, however, both ␣ 7BX1 -and ␣ 7BX2 -transfected cells may have reached saturation in adhesion to laminin-1. Furthermore, there is ample evidence that the cellular environment of ␤ 1 integrins greatly affects their affinity and specificity of binding. Thus, the interpretation of the role of distinct integrins in cell adhesion may be obscured by the presence of other integrins on the cell surface. For example, Ziober et al. (17) demonstrated that ␣ 7X1 -transfected MCF7 cells bound laminin-1 only after activation with the ␤ 1 -activating antibody, TS/16, whereas ␣ 7X1 -transfected HT1080 cells constitutively bound laminin-1. These data indicated a conformational dependence of the laminin binding site in the ␣ 7X1 extracellular domain, which can be regulated by ␤ 1 antibodies and by divalent cations, in particular Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ . Recently, antibody studies indicated laminin-induced conformational changes of ␣ 7 integrins in muscle cells (45).
␣ 7 ␤ 1 integrins bind only to native laminins and require the C-terminal portions of laminin ␣-, ␤-, and ␥-chains in their native trimeric conformation for binding (19,46,47). Since the laminin ␤ 1 and ␥ 1 chains are identical in laminins-1, -2, -8, and -10, differences in the laminin ␣-chains must account for the different affinities of the ␣ 7X1 and ␣ 7X2 variants for these laminins. There is only moderate sequence homology between the C-terminal portions of laminin ␣ 4 and ␣ 5 chains that comprise the globules of laminin-8 or -10/11, respectively. The degree of homology between laminin ␣ 4 and ␣ 5 is not higher than that between laminin ␣ 4 and ␣ 2 or ␣ 1 chains; it is therefore surprising that the ␣ 7X2 integrin recognizes laminin-1 and-2 but not laminin-8 and -10/11, while ␣ 7X1 recognizes all four laminins, although at different affinities. Neither ␣ 7X1 nor ␣ 7X2 bind to laminin-5, however, which contains the laminin ␣ 3 chain. Further studies on the integrin-binding epitopes of the laminin isoforms will be necessary for the characterization of the structural basis of specificity for interactions with integrins.
The question remains whether the absence of the cytoplasmic domains from the soluble complexes may affect the affinity and specificity of the extracellular domains to their ligands. There is no precedence, however, for a direct impact of the cytoplasmic domains of ␣-subunits on the ligand binding specificity or affinity of ␤ 1 integrins. The cytoplasmic domains of ␣-subunits are involved in the transduction of extracellular signals to the cytoskeleton and intracellular signaling pathways regulating cellular responses (48 -51). Deletion of the entire cytoplasmic domain of ␣ 7 integrin has been shown to have no effect on the adhesion of transfected HEK293 cells to laminins-1 or -2, but it does alter cell migration and phosphorylation reactions (52). It remains to be elucidated whether the laminin isoform-specific interactions of the soluble ␣ 7X1 ␤ 1 and ␣ 7X2 ␤ 1 variants reflect their affinities in vivo or whether they are modulated in vivo depending on the cell type in which they are expressed, similar to the cases of ␣ 7 ␤ 1 (17) or ␣ 2 ␤ 1 (53). Cell type-dependent modulation of binding affinities, however, was seen in tumor cells or after ectopic expression of integrins in cell lines. So far, cells expressing only the ␣ 7X1 splice variant have not been identified, but attempts are in progress to identify and to isolate such cells from muscle tissue to verify the affinity of the X1 splice variant for laminin-8 and -10.
The specific affinity of the ␣ 7X1 ␤ 1 variant for laminins-8 and -10/11 suggests that ␣ 7X1 may play a potential role in developing mouse skeletal muscle and also during situations associated with the reexpression of these laminins in regenerating muscle (21). ␣ 7 ␤ 1 integrin is also strongly expressed in smooth muscle that lines the walls of arterioles. In view of the widespread distribution of laminin-8 and -10/11 in embryonic muscle including the sarcolemma, neuromuscular junctions, and the endoneurium, it seems more likely that the ␣ 7X1 integrin splice variant is involved in heterologous cellular interactions in these locations. Therefore, the data presented may indicate that ␣ 7X1 ␤ 1 represents a potential receptor for laminins-8 and -10 during early embryogenesis, whereas ␣ 7X2 ␤ 1 in later stages