Integrin β Cytoplasmic Domains Differentially Bind to Cytoskeletal Proteins*

Integrin cytoplasmic domains connect these receptors to the cytoskeleton. Furthermore, integrin-cytoskeletal interactions involve ligand binding (occupancy) to the integrin extracellular domain and clustering of the integrin. To construct mimics of the cytoplasmic face of an occupied and clustered integrin, we fused the cytoplasmic domains of integrin β subunits to an N-terminal sequence containing four heptad repeat sequences. The heptad repeats form coiled coil dimers in which the cytoplasmic domains are parallel dimerized and held in an appropriate vertical stagger. In these mimics we found 1) that both conformation and protein binding properties are altered by insertion of Gly spacers C-terminal to the heptad repeat sequences; 2) that the cytoskeletal proteins talin and filamin are among the polypeptides that bind to the integrin β1A tail. Filamin, but not talin binding, is enhanced by the insertion of Gly spacers; 3) binding of both cytoskeletal proteins to β1A is direct and specific, since it occurs with purified talin and filamin and is inhibited in a point mutant (β1A(Y788A)) or in splice variants (β1B, β1C) known to disrupt cytoskeletal associations of β1 integrins; 4) that the muscle-specific splice variant, β1D, binds talin more tightly than β1A and is therefore predicted to form more stable cytoskeletal associations; and 5) that the β7 cytoplasmic domain binds filamin better than β1A. The structural specificity of these associations suggests that these mimics offer a useful approach for the analysis of the interactions and structure of the integrin cytoplasmic face.

Integrin cytoplasmic domains connect these receptors to the cytoskeleton. Furthermore, integrin-cytoskeletal interactions involve ligand binding (occupancy) to the integrin extracellular domain and clustering of the integrin. To construct mimics of the cytoplasmic face of an occupied and clustered integrin, we fused the cytoplasmic domains of integrin ␤ subunits to an N-terminal sequence containing four heptad repeat sequences. The heptad repeats form coiled coil dimers in which the cytoplasmic domains are parallel dimerized and held in an appropriate vertical stagger. In these mimics we found 1) that both conformation and protein binding properties are altered by insertion of Gly spacers Cterminal to the heptad repeat sequences; 2) that the cytoskeletal proteins talin and filamin are among the polypeptides that bind to the integrin ␤1A tail. Filamin, but not talin binding, is enhanced by the insertion of Gly spacers; 3) binding of both cytoskeletal proteins to ␤1A is direct and specific, since it occurs with purified talin and filamin and is inhibited in a point mutant (␤1A(Y788A)) or in splice variants (␤1B, ␤1C) known to disrupt cytoskeletal associations of ␤1 integrins; 4) that the muscle-specific splice variant, ␤1D, binds talin more tightly than ␤1A and is therefore predicted to form more stable cytoskeletal associations; and 5) that the ␤7 cytoplasmic domain binds filamin better than ␤1A. The structural specificity of these associations suggests that these mimics offer a useful approach for the analysis of the interactions and structure of the integrin cytoplasmic face.
Integrins, a large family of heterodimeric adhesion receptors, physically link the extracellular matrix to cytoskeletal elements within a cell (1)(2)(3). The structure of integrin cytoplasmic domains plays a key role in these linkages (4 -6). Thus, deletion or mutations in the cytoplasmic domain block the colocalization of integrins with cytoskeletal elements such as vinculin and talin (7,8). Furthermore, isolated ␤ cytoplasmic domains, joined to other transmembrane proteins, target them to focal adhesions, sites of integrin-cytoskeleton linkages (9,10). Thus, integrin ␤ cytoplasmic domains link these receptors to the cytoskeleton.
Integrin ␣ cytoplasmic domains limit certain cytoskeletal interactions of their ␤ partners. In intact integrins, focal adhesion targeting usually requires binding of a ligand to the extracellular domain (10), a conformational change in the extracellular domain of the integrin (11) and an intact ␤ tail (8).
Ligand-independent focal adhesion targeting is induced by deletion of the ␣ cytoplasmic domain (8,12), suggesting that the ␣ tail blocks cytoskeletal interaction with the ␤ tail. Ligand binding appears to remove this block, permitting the ␤ tail to target to focal adhesions. Consequently, isolated ␤ cytoplasmic domains resemble ligand-occupied integrins in their interactions with the cytoskeleton. The majority of integrin ligands are multivalent and cause receptor clustering after binding (13). Simple antibody-mediated clustering can initiate biochemical signals (14) and local accumulation of tensin and pp125 FAK near the clustered integrins (15). Nevertheless, accumulation of the full complement of cytoskeletal proteins (15) and functional cytoskeletal interaction (16,17) requires ligand occupancy. Consequently, cytoskeletal interactions of integrins involve both receptor clustering and ligand occupancy.
We previously proposed a strategy for the chemical synthesis of structural models of the cytoplasmic domain of multisubunit transmembrane receptors (18). The cytoplasmic domains of integrin ␣IIb␤3 were covalently linked via a helical coiled coil. In the present study we have extended and tested this approach by preparation of recombinant mimics of the cytoplasmic face of occupied and clustered integrins. By using recombinant proteins, we avoided limitations of polypeptide length and modest yield encountered in the initial synthetic approaches. Occupancy was mimicked by use of isolated ␤ cytoplasmic domains; clustering was mimicked by use of covalent homodimers of these domains. Helical coiled coil architecture provided the desired parallel topology and vertical stagger of the tails. We now report that the binding of cytoplasmic proteins to mimics of the ␤1A tail is sensitive to conformational changes induced by the insertion of Gly residues C-terminal to the helical coiled coil moieties. The cytoskeletal proteins talin and filamin are among the polypeptides that bind to the ␤1A tail. Filamin, but not talin binding, is enhanced by the insertion of a Gly spacer. Binding of both proteins is direct and specific, since it occurs with the purified proteins and is inhibited in a point mutant (␤1A(Y788A)) or in splice variants (␤1B (19) and ␤1C (20)) known to disrupt cytoskeletal associations of ␤1 integrins (21)(22)(23). The muscle-specific splice variant ␤1D (24 -26) binds talin better than ␤1A, and the ␤7 cytoplasmic domain binds filamin more tightly than ␤1A. Thus, these constructs can be used to analyze integrin class and splice variantspecific interactions with cytoskeletal proteins.
A human ␤1C cDNA was the gift of Dr. J. Meredith (The Scripps Research Institute, La Jolla) (21). ␤1A cDNA with the point mutation, Y788A, was kindly provided by Dr. A. F. Horwitz (University of Illinois, Champaign, Il) (22). A cDNA for the cytoplasmic domain of human integrin ␤1D (24) obtained by reverse transcription-PCR 1 of heart muscle total RNA (kindly provided by Dr. R. Ross (University of California, Los Angeles, CA)) was provided by Dr. J. Loftus (The Scripps Research Institute, La Jolla, CA). The cDNA of human integrin ␤7 has been described (27). A cDNA coding for the human ␤1B subunit cytoplasmic domain was synthesized in PCR reactions using a human ␤1A vector with a partially overlapping reverse oligonucleotide containing the published human ␤1B sequence (19).
Recombinant Cytoplasmic Domain Models-Based on an initial synthetic approach for the synthesis of models of the cytoplasmic domains of integrins (18), oligonucleotides were synthesized and used in PCR to create a cDNA encoding the four-heptad repeat protein sequence, KLE-ALEGRLDALEGKLEALEGKLDALEG (G1-([heptad] 4 ). Variants containing 1-3 additional Gly residues (G2-4-([heptad] 4 ) at the C terminus were synthesized by modification of the antisense oligonucleotide. These cDNAs were ligated into an NdeI-HindIII-restricted modified pET15b vector (Novagen). Integrin cytoplasmic domains were joined to the helix as HindIII-BamHI fragments. The final constructs coded for the N-terminal sequence GSSHHHHHHSSGLVPRGSHMCG[heptad] 4 linked to the cytoplasmic domains of integrins. Different integrin cytoplasmic domain cDNAs were cloned via PCR from appropriate cDNAs using forward oligonucleotides, introducing a 5Ј HindIII site, and reverse oligonucleotide, creating a 3Ј BamHI site directly after the stop codon. PCR products were first ligated into the pCR TM vector using the TA cloning ® kit (Invitrogen). After sequencing, HindIII/BamHI inserts were ligated into the modified pET15b vector. Recombinant expression in BL21(DE3)pLysS cells (Novagen) and purification of the recombinant products were performed according to the pET system manual (Novagen) with an additional final purification step on a reverse phase C18 high performance liquid chromatography column (Vydac). Products were analyzed by electrospray ionization mass spectrometry on an API-III quadrupole spectrometer (Sciex; Toronto, Ontario, Canada).
Ultraviolet Circular Dichroism Spectroscopy-Far UV CD spectra were recorded on an AVIV 60DS spectropolarimeter with peptides dissolved in 50 mM boric acid, pH 7.0. Data were corrected for the spectrum obtained with buffer only and related to protein concentrations determined from identical samples by quantitative amino acid analysis. From these values, the predicted percentage of helical secondary structure was calculated as described (18).
Cells and Cell Lysates-Human platelets were obtained by centrifugation of freshly drawn blood samples at 1000 rpm for 20 min and sedimentation of the resulting platelet-rich plasma at 2600 rpm for 15 min. They were washed twice with 0.12 M NaCl, 0.0129 M trisodium citrate, 0.03 M glucose, pH 6.5, and once in Hepes-saline (3.8 mM Hepes, 137 mM NaCl, 2.7 mM KCl, 5.6 mM D-glucose, 3.3 mM Na 2 HPO 4 , pH 7.3-7.4). Human Jurkat and HT1080 cells and mouse C2C12 cells were obtained from the American Type Culture Collection (Rockville, MD) and cultured in RPMI1680 (Jurkat) or Dulbecco's modified Eagle's medium with 10% fetal calf serum. For differentiation to myotubes, C2C12 myoblasts were kept confluent in Dulbecco's modified Eagle's medium with 5% horse serum for 6 days. Cultured cells were washed twice in phosphate-buffered saline and biotinylated with 1 mM Nhydroxysuccinimidobiotin-biotin (Pierce) in phosphate-buffered saline for 30 min at room temperature. Platelet proteins were biotinylated in Hepes-saline. After two additional washes with Tris-buffered saline, cells were lysed on ice with buffer A (1 mM Na 3 VO 4 , 50 mM NaF, 40 mM sodium pyrophosphate, 10 mM Pipes, 50 mM NaCl, 150 mM sucrose, pH 6.8) containing 1% Triton X-100, 0.5% sodium deoxycholate, 1 mM EDTA and protease inhibitors (20 g/ml) aprotinin, 5 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride. In addition, 0.1 mM E-64 (Boehringer Mannheim), a calpain inhibitor, was in the platelet lysate. Lysates were sonicated 5 times on ice for 10 s at a setting of 3 using an Astrason Ultrasonic Processor (Heart Systems; Farmingdale, NY). After 30 min, lysates were clarified by centrifugation at 12,000 g for 30 min.
Affinity Chromatography Experiments with Integrin Cytoplasmic Domain Mimics-500 g of purified recombinant cytoplasmic domain proteins were dissolved in a mixture of 5 ml of 20 mM Pipes, 50 mM NaCl, pH 6.8, and 1 ml of 0.1 M sodium acetate, pH 3.5, and bound overnight to 80 l of Ni 2ϩ -saturated His-bind resin (Novagen). In control experiments (data not shown) we found that these conditions resulted in saturation of the resin with recombinant protein. Resins were washed twice with 20 mM Pipes, 50 mM NaCl, pH 6.8, and stored at 4°C in an equal volume of this Pipes buffer containing 0.1% NaN 3 as a preservative. 50 l of such a suspension was added to 4.5 ml of cell lysates, which had been diluted 10-fold with buffer A containing 0.05% Triton X-100, 3 mM MgCl 2 , and protease-inhibitors. After incubation overnight at 4°C, resins were washed five times with this buffer and finally heated in 50 l of reducing sample buffer for SDS-PAGE. Samples were separated on 4 -20% SDS-polyacrylamide gels (NOVEX) and either stained with Coomassie Blue or transferred to Immobilon P membranes (Millipore). Membranes were blocked with Tris-buffered saline, 5% nonfat milk powder and stained with streptavidin peroxidase or specific antibodies. Bound peroxidase was detected with an enhanced chemiluminescence kit (Amersham Corp.). Equal loading of Ni 2ϩ resins with recombinant proteins was verified by Coomassie Blue staining of SDS-PAGE profiles of proteins eluted from the resins with SDS sample buffer.
Binding to Purified Talin and Filamin-Human uterus filamin was kindly provided by Dr. John Hartwig (Massachusetts General Hospital, Boston, MA) as a 1.5 mg/ml solution in 0.6 M KCl, 0.5 mM ATP, 0.5 mM dithiothreitol, 10 mM imidazole, pH 7.5. For binding assays performed as described above, this solution was diluted 1/12 with buffer A, 0.05% Triton X-100, 3 mM MgCl 2 , 2 mg/ml bovine serum albumin, and protease inhibitors (see above) (omitting the 50 mM NaCl (see above)), and resins with bound model proteins were added. Washing was performed in this buffer without bovine serum albumin and with additional 50 mM KCl.
Talin was provided by Dr. Keith Burridge (University of North Carolina) and was purified from human platelets essentially as described in (28) with an additional purification step using chromatography on phosphocellulose and stored at 1 mg/ml in 10 mM Tris acetate, pH 7.5, 0.5 mM EGTA, 0.5 mM EDTA, 0.05% ␤-mercaptoethanol, 100 mM NaCl, 50% glycerol. This solution was diluted to either 87 or 17 g/ml talin with buffer A, 0.05% Triton X-100, 3 mM MgCl 2 , 2 mg/ml bovine serum albumin, and protease inhibitors (see above, including 0.1 mM E-64) and processed as indicated in the binding assays with cell lysates. For densitometric analysis, scans of Coomassie-stained gels were processed using the program NIH-Image (NIH, Bethesda, MD).

RESULTS
Recombinant Integrin Cytoplasmic Domains-We developed a strategy for recombinant production of model proteins consisting of an N-terminal His tag sequence followed by a thrombin cleavage site, a Cys-Gly linker, and four heptad repeats connected to the integrin cytoplasmic domain (Fig. 1). These polypeptides are predicted to form parallel coiled coil dimers under physiological conditions (18). A cystine bridge ensures a parallel orientation and a correct stagger of the coiled coil sequences within the dimer. All synthesized proteins were Ͼ90% homogenous as judged by reverse phase C18 high pressure liquid chromatography ( Fig. 2A) and had a monomer mass that varied by less than 0.1% from that predicted for the desired sequence as judged by electrospray ionization mass spectrometry (Table I). The formation of covalent dimers in aqueous solution was observed by mass spectrometry (Fig. 2B) and by SDS-PAGE (Fig. 2C), thus confirming the parallel orientation of the helices.
The beginning of the integrin cytoplasmic domain sequence could provide the a and d residues of a fifth heptad repeat (Fig.  1). Consequently, direct linkage to the coiled coil sequence could induce helical structure in the tail. To address this possibility, we synthesized protein models containing Gly residues inserted between the coiled coil and the cytoplasmic domain sequence (Fig. 1). We compared the CD spectra of ␤1 integrin constructs containing either no (G1-␤1A) or three (G4-␤1A) additional Gly residues inserted between the coiled coil and the cytoplasmic domain. Insertion of glycines sharply reduced the minima at 208 and 222 nm. Consequently, predicted ␣-helical content in the protein model was reduced from 65 to 36% (Fig.  3). 29% of the amino acids in the construct are in the four heptad repeats; therefore, 36% helical content is consistent with most of the helical structure being limited to these repeats. Thus, the Gly insertion appears to eliminate ␣-helical structure induced in the cytoplasmic domain by the direct linkage to the coiled coil sequence.
Cytoplasmic Domain Conformation Affects Cytoskeletal Protein Binding-To assess functional consequences of the structural changes induced by the Gly insertions, we produced protein models of the ␤1A cytoplasmic domain with one, two, and three additional Gly residues inserted after the heptad repeat motif (G2-, G3-, G4-␤1A) and compared these with the G1-␤1A construct. As an additional control, we produced a variant of the G4-␤1A peptide, with a Tyr to Ala substitution in the membrane-proximal NPXY-motif (G4-␤1A(Y788A)) ( Fig. 1). This mutation interferes with focal adhesion targeting (22) and activation (29) of integrins. The purified proteins were bound via their N-terminal His tag to a Ni 2ϩ resin and used in affinity chromatography experiments with lysates of human platelets. Marked changes in protein binding properties were observed as a consequence of the Gly insertions ( Fig. 4): In Coomassie Blue-stained gels, polypeptides migrating at 45, 56, 58, 140, and 240 kDa bound only to the mimics with Gly insertions. Moreover, comparison to the cell lysate (Ly in Fig. 4A) showed that this binding was selective. Immunoblotting identified the 240-kDa and 45-kDa proteins as filamin and actin, respectively (Fig. 5). The enriched 56-, 58-, and 140-kDa polypeptides have not been identified. They failed to react with antibodies specific for pp60 src (Fig. 5), paxillin, pp125 FAK , ␣-actinin, and vinculin (data not shown) in immunoblotting experiments. Talin bound to both the G1-and G4-␤1A protein. In addition, we used biotinylated extracts to enhance sensitivity of protein detection. In such extracts, we confirmed talin binding to the G1-␤1A protein and enhanced filamin binding with the Gly insertions (Fig. 4B). Thus, the structural changes (Fig. 3) induced by the insertion of Gly between the coiled coil motif and the integrin cytoplasmic domain sequence alter interactions of these proteins with cellular components.
To establish specificity of the binding of cellular proteins to G4-␤1A, we examined the effect of structural changes known to inhibit cytoskeletal associations of the intact integrin. Introduction of the Y788A mutation into the G4-␤1A construct (YA) suppressed the interaction with both talin and filamin but not the binding of the other polypeptides (Fig. 4). Selective suppression of the binding of talin and filamin was also observed with G4-␤1B and G4-␤1C constructs (data not shown, but see Fig. 8). Thus, alterations of the ␤1A tail, which block cytoskeletal interactions, also reduce binding to talin and filamin.
The ␤1A cytoplasmic domain is dimerized in these model proteins. To assess the role of ␤1A dimerization in cytoskeletal protein binding, we constructed a heterodimer containing a single G4-␤1A subunit. The other subunit (G4-␣T) was identical to G4-␤1A through the four glycines. C-terminal to the glycines, we placed a KLGFFKR sequence, representing the conserved membrane proximal region of integrin ␣ cytoplasmic domains (4). The structure of the heterodimer was confirmed by mass spectroscopy (Table I). The G4-␣T␤1A heterodimer bound filamin and talin to the same extent as the G4␤1A dimer (Fig. 6). Specificity of binding in this experiment was confirmed by the failure of an ␣IIb cytoplasmic domain construct to bind either cytoskeletal protein. Thus, the dimerization of the ␤1A cytoplasmic domain is not essential for talin and filamin binding.
Differential Binding of Talin and Filamin to Integrin ␤1D and ␤7 Cytoplasmic Domains-To extend our affinity chromatography results, we prepared similar G1 and G4 model proteins of the splice variant, ␤1D, and the ␤7 integrin subunits (Fig. 1). When incubated with platelet lysates, the ␤1D constructs bound more talin, and ␤7 constructs bound more filamin compared with ␤1A (Fig. 7A). In addition, these differences in binding were consistently observed when lysates of a human T-cell leukemia cell line (Jurkat), a human fibrosarcoma cell line (HT 1080), or differentiated myotubes derived from a mouse myoblast cell line (C2C12) were used for affinity chromatography (Fig. 7B). Moreover, stronger binding of the ␤1D constructs to talin and of the ␤7 constructs to filamin was observed both with the G1 (data not shown) as well as the G4 variants of the model proteins, indicating that structural  Fig. 4 with antibodies specific for filamin, talin, actin,  and pp60 src . Ly, lysate; Co, unloaded Ni 2ϩ resin; YA, G4-␤1A(Y788A).
FIG. 6. Role of ␤1A dimerization in talin and filamin binding. N 1 2ϩ resin was loaded with the heterodimer G4-␣T␤ 1 A or homodimers of G4-␤ 1 A or G4-␣ IIb . Bound proteins from platelet extracts were separated on 4 -20% SDS-polyacrylamide gels under reducing conditions, transferred to nitrocellulose membranes, and stained with antibodies specific for talin or filamin. In addition, Coomassie Blue-stained gels were examined to assess loading of the resin. Note the migration position of the G4-␣T construct (␣T) and the similar quantities of G4-␤1A or G4-␣ IIb subunit in each lane. changes induced by Gly insertions do not alter the relative strengths of these interactions.
Interactions with Purified Talin and Filamin-To learn whether the observed interactions with talin and filamin in the cell extracts are direct, we used purified preparations of these proteins. The relative binding of purified filamin and talin to the model proteins was similar to that observed with cell lysates (Fig. 8). Specifically, ␤1D constructs bound more talin, and ␤7 constructs bound more filamin than ␤1A (Fig. 8). In addition, binding of both cytoskeletal proteins to the G4-␤1A(Y788A) construct and to the G4-␤1B and G4-␤1C variants was markedly reduced compared with G4-␤1A (Fig. 8). Moreover, G4 constructs of ␤1A, ␤1D, and ␤7 integrin cytoplasmic domains bound more purified filamin than the corresponding G1 constructs. However, the G1-␤7 model protein still bound more filamin than G4-␤1A or G4-␤1D (data not shown). A densitometric analysis of the Coomassie Blue-stained proteins indicated that the ␤1D construct bound 9 times more talin, and the ␤7 construct bound 8.4 times more filamin than the ␤1A model protein (Fig. 8). Thus, filamin and talin bind differentially to the different integrin ␤ subunit cytoplasmic domain constructs. DISCUSSION To study integrin-cytoskeletal interactions, we designed structural mimics of the cytoplasmic face of occupied and clustered integrins. We joined integrin cytoplasmic domains to an N-terminal parallel coiled coil dimer, and we found that the introduction of Gly spacers between the coiled coil and the ␤1A integrin tail leads to a change in the conformation of the protein. Concomitantly, the polypeptide binding properties are altered, resulting in an increase in the strength of the interaction with the cytoskeletal protein, filamin. Talin and filamin binding to these structural mimics are direct, specific, and biologically relevant. These interactions are observed with purified talin and filamin and are abrogated in splice variants and a mutant that disrupt cytoskeletal interactions of the intact integrin. The ␤1D integrin tail binds talin, and the ␤7 integrin tail binds filamin much more tightly than the ␤1A cytoplasmic domain. Thus, we provide insight into integrin class and splice variant-specific interactions with cytoskeletal proteins and their dependence on the conformation of integrin ␤ cytoplasmic domains.
The model proteins described here have the topology of clustered integrin cytoplasmic domains. The tropomyosin-derived heptad repeats employed (30) are known to form parallel dimers (31,32). Moreover, they were dimeric and disulfidebonded (Fig. 2C) via their N termini. Thus, they must have been parallel, and the integrin tails must have been in the desired vertical stagger. The conformation of these constructs was sensitive to Gly insertions between the coiled coil and the integrin tails. The coiled coil structure formed by the heptad repeats should not be disrupted by these Gly insertions, and the estimated 36% residual helix in the G4-␤1A construct is consistent with maintenance of the coiled coil. Consequently, the Gly insertions may have disrupted induced helical structure in the ␤1A tail in the G1-␤1A construct.
These model proteins permitted direct detection of talin and filamin binding to integrin ␤1A subunit cytoplasmic domains. Binding of both talin and filamin was specific since it was sensitive to alterations in the ␤ tail sequence known to block cytoskeletal associations in vivo. Previously, equilibrium gel filtration was required to demonstrate low affinity ␤1A integrin-talin interactions (33). Furthermore, affinity chromatography or yeast two-hybrid screens with linear ␤1A cytoplasmic domain peptides failed to isolate either filamin or talin (34 -37). The recombinant model proteins present the tails as vertically oriented dimers. Furthermore, they are immobilized via an N-terminal His tag. The coiled coil also serves to introduce an additional ϳ42-Å space (38) between the immobilizing surface and the tail. However, we found that a monomeric ␤ tail, presented on a similar model protein, also bound talin and filamin. The vertical orientation of the ␤ tails and the reduction of potential steric hindrance may also contribute to the efficiency of these model proteins in affinity chromatography.
The differential binding of ␤ cytoplasmic domains to talin and filamin suggests functionally different linkages of integrins with the cytoskeleton. Talin and non-muscle filamin are both reported to be actin-binding proteins (3), but they seem to link to different components of the cytoskeleton. Non-muscle filamin cross-links actin filaments and promotes their high angle branching (39). It is thus part of the cortical actin network. The preferential binding of the integrin ␤7 tail to filamin could therefore affect the cell surface distribution of ␤7 integrins and contribute their role in lymphocyte homing (40,41). In contrast, talin concentrates at sites in which integrins are clustered, and actin filaments insert end-on into submembraneous focal complexes (2,3). Such sites include tension-bearing structures such as myotendinous junctions, intercalated disks, and costameres in striated muscle (42). The markedly increased binding of the muscle-specific splice variant, ␤1D, to talin could provide a molecular mechanism for increased stability of matrix-cytoskeletal linkages at these sites.
The model proteins described here provide new approaches to the structure and function of integrin cytoplasmic domains. We found that talin binding was increased in the ␤1D splice variant, yet the major differences in amino acid sequence between ␤1D and ␤1A lie outside of the previously identified talin binding site (43) in ␤1A. Furthermore, although filamin had not been noted to bind ␤1A, it was reported to bind to the ␤2 cytoplasmic domain in affinity chromatography experiments (44). In the previous work, the filamin binding site was localized to a relatively conserved membrane-proximal REYR-RFEKEK sequence. In our experiments, filamin and talin binding were sensitive to changes at membrane-distal sites in the ␤1B and ␤1C splice variants and the ␤1A(Y788A) mutant. Taken together, these data suggest that the folded structure of the ␤ cytoplasmic domain is important in its interactions with cytoskeletal proteins. The G4-␤1A and G1-␤1A constructs contain identical ␤1A sequences. Their marked differences in cytoskeletal protein binding represents direct evidence of the functional importance of the three-dimensional structures of integrin ␤ cytoplasmic domains. These model proteins may also provide powerful tools for the elucidation of such structures.