Identification of the region in actin-binding protein that binds to the cytoplasmic domain of glycoprotein IBalpha.

Actin-binding protein (ABP-280) is a component of the submembranous cytoskeleton and interacts with the glycoprotein (GP) Ibα subunit of the GP Ib-IX complex in platelets. In the present studies, we have identified the binding site for GP Ibα in ABP-280. A melanoma cell line lacking ABP-280 was stably transfected with the cDNAs coding for GP Ib-IX, then transiently transfected with cDNA coding for various carboxyl-truncates of ABP-280. Immunocapture assays and co-immunoprecipitation experiments from detergent-lysed cells showed that deletion of the carboxyl-terminal repeats 20-24 of ABP-280 had no effect on GP Ib-IX binding, but deletion of residues 2099 through 2136 within repeat 19 abolished binding. In the yeast two-hybrid system, an ABP-280 fragment comprising repeats 17-19 bound GP Ibα. Deletion from either end abolished binding. Individual or multiple repeats of ABP-280 were expressed as fusion protein in bacteria and purified; structural folding was evaluated, and binding to GP Ib-IX was assessed. Binding depended on the presence of repeats 17-19. None of the individual repeats were able to bind to GP Ib-IX. These findings demonstrate that residues 1850-2136 comprising repeats 17-19 contain the binding site for GP Ib-IX.

The activation and association of membrane receptors with the submembranous cytoskeleton is of key importance in regulating cellular functions such as adhesion, motility, transmembrane signaling, receptor distribution, and receptor function. A major component of the submembranous cytoskeleton is actin-binding protein-280 (ABP-280), 1 also known as filamin. In platelets, ABP-280 cross-links actin filaments and anchors the membrane skeleton to the plasma membrane by interacting with the cytoplasmic tail of the von Willebrand factor receptor, the glycoprotein (GP) Ib-IX complex (1)(2)(3)(4)(5)(6). GP Ib-IX is a heterotrimeric complex consisting of two disulfide-linked subunits, GP Ib ␣ (M r 140,000) and GP Ib ␤ (M r 25,000), that are noncovalently associated with GP IX (M r 22,000) (7,8). These glycoproteins are each transmembrane proteins (9 -11). The interaction of the GP Ib-IX complex with ABP-280 is mediated by a region in the central portion of the cytoplasmic domain of GP Ib ␣ (5,6,12). In other cells of myeloid lineage that do not express GP Ib-IX, ABP-280 binds the high affinity IgG receptor (Fc␥RI) and the ␤2 integrin (13,14).
Biochemical and structural analysis of human ABP-280 and cloning of the human endothelial ABP-280 cDNA reveals a protein of 2647 amino acids with three functional domains (15). ABP-280 subunits self-associate head to head (16) using the most carboxyl repetitive element (15,17). The opposing aminoterminal end of each subunit contains an actin-binding domain (15,(17)(18)(19). The bulk of ABP-280 forms a semiflexible rod domain and is composed of 24 repeats, each about 96 residues long, that are predicted to fold into 8 ␤-sheets. The rod domain of ABP-280 is interrupted twice by short sequence inserts of 20 -40 residues between repeats 15 and 16 and repeats 23 and 24. These regions are postulated to be flexible hinges, and they both contain a calpain cleavage site. Calpain acts initially at the region between repeats 15 and 16 to generate an aminoterminal fragment of 190 kDa and a carboxyl-terminal fragment of 100 kDa (20). The 100-kDa fragment is subsequently cleaved at the site between repeats 23 and 24 to generate 90and 10-kDa subfragments. Previous reports showed that in platelet lysates in which ABP-280 has been cleaved, GP Ib-IX co-immunoprecipitates with the 100-kDa fragment and to a lesser extent with the 90-kDa fragment (4,6,21), suggesting that the binding site for GP Ib ␣ is located in the carboxylterminal portion of ABP-280.
In the present studies, using three different approaches, we characterized the region of ABP-280 that interacts with GP Ib ␣ . In the first approach, a melanoma cell line that lacks ABP-280 (22) was stably transfected with the cDNAs coding for the three subunits of the GP Ib-IX complex, then transiently transfected with cDNAs coding for ABP-280 lacking increasing numbers of amino acids from the carboxyl-terminal end; the ability of the truncates to associate with GP Ib-IX was determined. An ABP-280 fragment truncated at residue 2136 at the border of repeats 19 and 20 (yielding a fragment lacking 511 carboxyl-terminal residues) retained GP Ib ␣ binding. In contrast, a fragment truncated at residue 2099 failed to bind. These results suggest that repeats 20 -24 are not required for binding of GP Ib ␣ but that removal of residues 2099 to 2136 of the carboxyl-terminal third of repeat 19 abolishes binding. As a second approach, the yeast two-hybrid system was used (23)(24)(25)(26)(27). In this system, an ABP-280 fragment of 286 amino acids, spanning residues 1850 -2136 comprising repeats 17-19, bound GP Ib ␣ . Fragments with deletions at either end of this fragment were unable to bind. Lastly, repeats 17-19 were expressed in bacteria individually or as multiple repeats. Repeats expressed individually or in tandem were unable to bind to GP Ib-IX; as in the two-hybrid system, a fragment consisting of repeats 17-19 (residues 1850 -2136) was required for binding to GP Ib ␣ .

Characterization of the Binding Domain by the Two-hybrid System-
Yeast plasmids containing the GAL4 DNA binding domain (pGBT9) (26) and the GAL4 activation domain (pGAD424) (26) were obtained from Clontech Laboratories (Palo Alto, CA). The yeast strain SFY526 (28) (Clontech) was used to assay protein-protein interactions. All yeast manipulations as well as ␤-galactosidase enzyme activity assays were performed as recommended by the supplier (Clontech).
Expression of ABP-280 Fragments in Melanoma Cells-The melanoma cells used in these studies were from the human melanoma cell line M2 lacking ABP-280 (22). The cells were stably transfected with the cDNAs encoding GP Ib ␣ , GP Ib ␤ , and GP IX, as described previously (29). Subsequently, cells were transiently transfected by the calcium phosphate method (30) with the cDNA coding for either full-length or truncated ABP-280. The cells were harvested 48 h after transfection.
Co-immunoprecipitation of ABP-280 with GP Ib-IX-To determine whether ABP-280 fragments expressed in melanoma cells associated with GP Ib-IX, cells were harvested with EDTA, and lysed in a Triton X-100-containing buffer; the detergent-soluble fractions were isolated by centrifugation as described previously (29). Monoclonal antibody ABP-4 (MAb ABP-4) (4, 15) or mouse IgG (Sigma Immuno Chemical Co.) were covalently coupled to 10 M polyacrylamide immunobeads (Bio-Rad) according to the manufacturer's instructions. After preclearing with mouse IgG-coupled immunobeads, the Triton X-100-soluble fractions were incubated with MAb ABP-4-coated beads. After extensive washing, the immunoprecipitated proteins were removed from the beads and analyzed on Western blots (29). The blots were first incubated with a monoclonal antibody against GP Ib ␣ (MAb Ib-4, courtesy of Dr. B. Steiner) (32) and then reprobed with a monoclonal antibody against ABP-280 (MAb ABP-4) (4, 15). Antigen-antibody complexes were detected by enhanced chemiluminescence using Amersham's ECL system.
To determine whether ABP-280 fragments expressed in bacteria associated with GP Ib-IX, ABP-deficient melanoma cells, expressing GP Ib-IX, were lysed as described above. Triton X-100-soluble fractions were incubated for 5 h with GST fusion protein-agarose bead complexes (Sigma).
Expression of ABP-280 Repeats as Fusion Protein in Bacteria-ABP-280 cDNAs corresponding to the desired amino acids were generated by PCR and ligated into a pGEX vector (Pharmacia Biotech Inc.). All 3Ј primers contained stop codons in all three reading frames to insure proper termination of translation. The sequences of the amplified fragments were verified by DNA sequencing analysis.
Fusion proteins were expressed in Escherichia coli. Translation of the vector was induced with 1 mM isopropyl ␤-thio-galactopyranoside for 3 h. Cells were harvested by centrifugation at 6600 ϫ g for 10 min and lysed by sonication; the resulting supernatant was passed over a glutathione-agarose column. Before evaluation of the secondary structure, GST was cleaved by thrombin (Sigma) (25 units/liter of bacterial culture). Released proteins were collected by centrifugation at 2500 ϫ g for 5 min, and purity was evaluated by SDS-polyacrylamide gel electrophoresis. Protein concentrations were determined using the BCA assay (Pierce).
Evaluation of Secondary Structure by Circular Dichroism-Polypeptides expressed in bacteria at final concentrations of 1 mg/ml were extensively dialyzed in 140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 1.8 mM KH 2 PO 4 , pH 7.5, and brought to room temperature before spectral analysis. Measurements were performed in an Aviv 60DS spectropolarimeter at room temperature. Cuvette path lengths of both 1 and 0.1 cm were used, and measurements were taken at step sizes of 0.5 nm. Secondary structure was determined using the software program Theta, which correlates spectra and predicted secondary structure.

Binding of ABP-280 to GP Ib-IX in Melanoma Cells Does Not
Require Repeats 20 -24 of the ABP-280 Subunit-As a first approach to identifying the region of ABP-280 that interacts with GP Ib ␣ , we used the melanoma cell line M2 deficient in ABP-280 (22). M2 cells were stably transfected with the cDNAs encoding GP Ib ␣ , GP Ib ␤ , and GP IX. Even in the absence of ABP-280, the three subunits were expressed, formed a complex, and were inserted into the plasma membrane. Fig. 1 demonstrates the surface expression of GP Ib-IX in cells stained with an antibody against GP Ib ␣ . Similar results were obtained with an antibody against GP IX (data not shown).
The GP Ib-IX-expressing cells were then transiently transfected with the constructs shown in Fig. 2 containing cDNAs coding for ABP-280 lacking increasing numbers of amino acids from the carboxyl-terminal end. The amount of ABP-280 carboxyl truncates expressed in the cells transfected with the different constructs was very similar (data not shown). To determine whether GP Ib-IX associated with the various truncated forms of ABP-280, the transfected cells were lysed in a Triton X-100-containing buffer. The ABP-280 fragments were captured with an anti-ABP antibody absorbed onto microtiter plates, and the amount of GP Ib-IX associated with the fragments was determined. Fig. 3A shows that like full-length ABP-280 in stably transfected cells (construct 1), ABP-280 truncates expressed in cells transiently transfected with constructs 2-5 (lacking 27 to 257 carboxyl-amino acids contained within the last third of repeat 22 through repeat 24) were recovered in association with GP Ib-IX. Similarly, a protein lacking repeats 20 -24 (deletion of 511 carboxyl-terminal amino acids), expressed in cells transiently transfected with construct 6, was able to bind to GP Ib-IX just as well as full-length ABP-280 (Fig. 3B). In contrast, an ABP-280 fragment missing 548 amino acids (lacking repeats 20 -24 plus the carboxylterminal third of repeat 19), expressed just as efficiently as the fragment lacking 511 carboxyl-terminal amino acids (Fig. 3,  inset, compare lanes 7 and 6), did not bind to GP Ib-IX (Fig. 3B,  column 7). Similar results were obtained when GP Ib ␣ was captured with an anti-GP Ib ␣ antibody and the amount of ABP-280 associated was measured (data not shown).
These results suggest that repeats 20 -24 of ABP-280 are not required for binding of ABP-280 to GP Ib-IX, but that residues 2099 -2136 in the carboxyl-third of repeat 19 are essential.
Further Characterization of the GP Ib ␣ Binding Site in the Yeast Two-hybrid System-As a second approach to identifying the GP Ib-IX binding site on ABP-280, we used the two-hybrid system (23)(24)(25)(26)(27). Based on previous findings, suggesting that the binding site for GP Ib ␣ was located in the carboxyl-terminal third of ABP-280 (4,21), the cDNA encoding the carboxyl-third of ABP-280 (aa 1585-2647, consisting of the middle of repeat 14 through repeat 24) was fused to the GAL4 activation domain. The cDNA encoding the cytoplasmic domain of GP Ib ␣ was fused to the GAL4 DNA binding domain. The recombinant plasmids were co-transformed into the SFY526 yeast strain. The interaction between GP Ib ␣ and ABP-280 reconstituted the function of GAL4; ␤-galactosidase activity was induced, resulting in the generation of blue colonies (Fig. 5). Neither fusion protein alone was able to activate the transcription of the reporter gene. These results confirmed that the binding site for GP Ib ␣ was contained within the carboxyl-terminal third of the molecule. As in the melanoma cells, deletion of the last 511 carboxyl-terminal amino acids of the ABP-280 subunit (aa 1585-2136, consisting of the middle of repeat 14 through repeat 19) did not affect its binding to GP Ib ␣ in the two-hybrid system (Fig. 5). Binding was lost, however, when just 37 more residues were deleted (deletion of the last 548 carboxyl-terminal amino acids).
The results presented thus far demonstrate that a fragment of ABP-280 consisting of the middle of repeat 14 through repeat 19 (residues 1585-2136) can bind GP Ib ␣ . To determine which part of the amino-terminal portion is required for binding, constructs containing cDNA coding for ABP-280 fragments with increasing deletions from the amino-terminal end were prepared. A truncated fragment of 286 amino acids containing residues 1850 -2136 (comprising repeats [17][18][19] was still able to bind to GP Ib ␣ (Fig. 5). However, further deletions of this fragment from its amino-end abolished the binding to GP Ib ␣ (Fig. 5); a truncate consisting of residues 1950 -2136 within repeats 18 -19 failed to bind to GP Ib ␣ . When a truncate consisting of residues 1850 -2136 of ABP-280 was expressed, blue colonies could be detected after 30 min, whereas a 15-18-h incubation was necessary to detect an interaction with a trun- cate composed of residues 1585-2136. With a liquid assay, 166 ␤-galactosidase units were generated with the shorter construct compared to 3.4 with the longer construct (Fig. 5).
Repeats 17-19 of ABP-280 Expressed as Fusion Proteins in Bacteria Bind to GP Ib-IX-It is conceivable that the shorter fragments of ABP-280 expressed in yeast (aa 1950 -2136 and aa 2020 -2136) may have lost their ability to bind to GP Ib-IX because of loss of their secondary structure. Thus, an additional assay using GST-fusion proteins was used. Individual repeats 18 (aa 1950 -2038), 19 (aa 2039 -2135), and 19 ϩ 32 aa (aa 2039 -2167), or multiple repetitive elements 18 -19 (aa 1950 -2135) were expressed as GST-fusion proteins in bacteria and purified; their structural folding was evaluated by circular dichroism. Human ABP-280, a protein composed primarily of ␤-structure, has a single minimum wavelength of 218 nm (33). In the present study, evaluation by the Theta algorithmic program indicated that purified human ABP-280 had approximately 13% ␣-helix, 62% ␤-sheet, and 25% random coil. Purified polypeptides also had a minimum at 218 nm, and the percentage of ␣-helix was between 11 and 12%, ␤-sheet was between 59 and 62%, and random coil was between 28 and 30%, suggesting that the purified recombinant repeats achieved secondary structure comparable to that of purified human ABP-280.
Purified GST fusion proteins coupled to agarose beads were incubated with extracts from Triton X-100-lysed, ABP-deficient melanoma cells expressing GP Ib-IX. GST beads alone (Fig. 6, lane 2) did not bind GP Ib ␣ ; neither did repeat 18 alone (Fig. 6,  lane 3), repeat 19 alone (Fig. 6, lane 4), the tandem repeats 18 -19 (Fig. 6, lane 5), or repeat 19 ϩ 32 aa (Fig. 6, lane 6). In contrast, repeats 17-19 expressed as GST-fusion protein (Fig.  6, lane 7) bound GP Ib ␣ . These results confirm the results obtained with the two-hybrid system that the binding site for GP Ib ␣ on ABP-280 is located in repeats 17-19 between residues 1850 and 2136. DISCUSSION The goal of the present study was to identify the domain on ABP-280 that binds GP Ib ␣ . Three different approaches have lead to the same conclusion that the binding site for GP Ib ␣ is located within repeats 17-19 of ABP-280. In the first approach, melanoma cells deficient in ABP-280 but expressing GP Ib-IX were transfected with various cDNAs directing the synthesis of carboxyl-deleted ABP-280. ABP-280 truncates lacking carboxyl-terminal residues from 27 to 511 associated with GP Ib-IX in these cells, but further deletion of 37 residues abrogated GP Immunocapture assays showing which of the ABP-280 truncates associate with GP Ib-IX. The GP Ib-IX-expressing cells, deficient in ABP-280, were transiently transfected with one of the six cDNAs coding for carboxyl termini ABP-280 truncates (shown in Fig.  2). The cells were harvested 48 h after transfection and lysed in a Triton X-100containing lysis buffer (see "Materials and Methods"). ABP-280 fragments from 25,000 to 160,000 cells were captured with an anti-ABP-280 antibody, and the amount of GP Ib-IX associated with ABP-280 was determined. The numbers on the top of each column correspond to the expressed ABP-280 carboxyl truncates as defined in Fig. 2. The unlabeled columns represent the signal obtained with cells deficient in ABP-280, and the columns la- Ib-IX binding. The deletion of the carboxyl-third of the repeat 19 (residues 2099 -2136), therefore, disrupts GP Ib ␣ binding to ABP-280. A second approach used the yeast two-hybrid system (23)(24)(25)(26)(27). Portions of the carboxyl terminus of ABP-280 were co-expressed along with the cytoplasmic tail of GP Ib ␣ . The smallest fragment of ABP-280 that bound GP Ib ␣ was composed of repeats 17-19 (residues 1850 -2136). Further deletions from either end of this fragment abolished the binding to GP Ib ␣ . These results demonstrate that the GP Ib ␣ binding site is near the middle of the 90 kDa rod fragment. These results were confirmed by expressing individual or multiple ABP-280 repeats as fusion protein in bacteria and testing their binding to GP Ib-IX. Repeats 17-19 (residues 1850 -2136), but not repeat 18 or 19 or the tandem 18 -19, bound to GP Ib-IX. Circular dichroism measurements indicated that the folding of the expressed proteins was comparable to that in native ABP-280. The combination of these three approaches strongly suggests that the binding site on ABP-280 for GP Ib ␣ is contained in repeats 17-19 (residues 1850 -2136). This is somewhat surprising because a great degree of conformity to the consensus repeat exists in repeats 17 and 19, and we would expect the binding site to be located in a less conserved region. A bigger divergence exists in repeat 18, where the regular pattern is interrupted by a 9-residue deletion at the initial portion of the repeat. This divergence from the consensus sequence may play a role in the binding of GP Ib ␣ .
The rod domain of ABP-280 is interrupted at two points by sequence inserts that contain a calpain cleavage site. Earlier studies showed that in platelet lysates in which ABP-280 was cleaved by calpain, GP Ib-IX remained associated with the 100-kDa carboxyl portion of the polypeptide chain and to a lesser extent with its 90-kDa calpain-generated subfragment (4,6,21). This suggested that GP Ib-IX might bind near the calpain cleavage site in the hinge just before repeat 24 (15). Alternatively, calpain cleavage could perturb the conformation of the 90-kDa subfragment and thereby diminish GP Ib-IX binding. The location of the GP Ib ␣ binding site determined in the present study suggests the latter mechanism to explain the decreased amount of GP Ib ␣ associated with the 90-kDa subfragment.
ABP-280 has proved difficult to dissociate from the GP Ib-IX complex in detergent-lysed platelets (1)(2)(3)(4)34). This observation stands in marked contrast to the 1 ϫ 10 Ϫ7 M Ϫ1 affinity of ABP-280 for GP Ib-IX measured in vitro (5). An experimental observation may shed some light on this discrepancy. The interaction between the fragment of ABP-280 formed by repeats 17-19 (residues 1850 -2136) and GP Ib ␣ was relatively strong in the twohybrid system compared to the one obtained with larger fragments from ABP-280 (Fig. 5). This might suggest that binding to GP Ib ␣ can be modulated by flanking sequence on both ends of repeats [17][18][19]. Many other factors could, of course, contribute to the weaker signal obtained in the two-hybrid system with larger fragments; for example, these fragments may be folded inappropriately, they may be less stable than the shorter ones, they may be expressed less efficiently than the smaller fragments, or there FIG. 5. Truncated ABP-280 fusion proteins that reconstituted or failed to reconstitute the ␤-galactosidase activity in yeast. The cDNA coding for the cytoplasmic domain of GP Ib ␣ was fused to the GAL4 DNA binding domain, and the cDNAs coding for truncated forms of ABP-280 were fused to the GAL4 activation domain. pGBT9-GP Ib ␣ and each activation domain fusion were co-transformed into the SFY526 yeast strain. Horizontal bars represent the amino acid (aa) sequence of the entire ABP-280 and the various truncated forms of ABP-280 expressed as fusion proteins in yeast. The numbers on the left of the horizontal bars indicate the N-terminal amino acid, and the numbers on the right indicate the carboxyl-terminal amino acid. The numbers in the horizontal bars correspond to the repeat number. The repeats were not numbered when truncated. The ability of each fusion protein to activate ␤-galactosidase activity, as assayed by filter and liquid assays, is indicated to the right of the figure. For the filter assays (39), positives appeared in 15-18 h, except for residues 1850 -2136, which appeared in 30 min. For the liquid assays, the absorbance of the samples was determined after a 5-h incubation, except for the last four values, where the absorbance was determined after 10 min. The numbers represent the ␤-galactosidase activity in Miller units (40). Values reported are the average of duplicate assays of three independent transformants. ND, not determined.
FIG. 6. Co-immunoprecipitation of GP Ib ␣ with ABP-280 repeats expressed as fusion protein in bacteria. ABP-deficient melanoma cells expressing GP Ib-IX were lysed in a Triton X-100-containing buffer (see "Materials and Methods"). The Triton-soluble fractions were incubated with various purified GST fusion proteins coupled to agarose beads. Immunoprecipitated proteins were electrophoresed through a SDS-polyacrylamide gel under reducing conditions and then transferred to nitrocellulose. The blot was probed with an anti-GP Ib ␣ antibody (MAb Ib-4 may be differences in the accessibility of the interacting domains to each other or accessibility of the activation domain to the transcriptional machinery. Human GP Ib-IX has been expressed in hamster and mouse nonhematopoietic cells in a form that associated with ABP-280 (35). Thus, the domain on ABP-280 that associates with the cytoplasmic domain of GP Ib ␣ appears to have been conserved. ABP-280 is a prominent component of many different cell types (36,37), and it now appears that the rod of this protein may act as a scaffolding to collect and tether diverse molecules at the cytoplasmic surface of the plasma membrane. Membrane association is conferred in platelets and megakaryocytes by the binding of repeats 17-19 to GP Ib-IX complexes. Other membrane receptors that bind to ABP-280 are the IgG Fc receptor I (Fc␥RI) (13) and integrin ␤ 2 (14). Sequence comparison between the cytoplasmic tails of GP Ib ␣ , Fc␥RI, and integrin ␤ 2 has, however, revealed no sequence conservation (9,14,38). Thus, the domain on ABP-280 implicated in the interaction with Fc␥RI and integrin ␤ 2 is likely to be different from the one interacting with GP Ib ␣ . Other proteins now known to bind along ABP-280 include small GRP-binding proteins, the endosomal endoproteinase furin, 2 and the transcription factor ATFa. 3 Future studies will be needed to characterize the way in which ABP-280 interacts with additional membrane glycoproteins and to elucidate the importance of ABP-280 in regulating the properties of these plasma membrane receptors.