A Conserved Sequence Motif in the Integrin β3 Cytoplasmic Domain Is Required for Its Specific Interaction with β3-Endonexin

Integrin signaling is mediated by interaction of integrin cytoplasmic domains with intracellular signaling molecules. Recently, we identified a novel 111-amino acid polypeptide, termed β3-endonexin, which interacts selectively with the integrin β3 cytoplasmic domain. In the present study we conducted a systematic mutational analysis of both the integrin β3 cytoplasmic domain and β3-endonexin to map sites required for interaction. The interaction of the full-length β3 integrin subunit with β3-endonexin in vitro required the β3 cytoplasmic domain. In a yeast two-hybrid system, both membrane-proximal and membrane-distal residues of the β3 cytoplasmic domain were necessary for interaction with β3-endonexin. In particular, the membrane-distal NITY motif at β3 756-759 was critical for the interaction. Exchange of β3 residues 756-759 (NITY) for the corresponding residues in β1 (NPKY) endowed the β1 cytoplasmic domain with the ability to interact with β3-endonexin. Conversely, exchange of the NPKY motif at β1 772-775 for the NITY motif in β3 abolished interaction of this chimeric cytoplasmic domain with β3-endonexin. Because the NITY motif is present in the β3 but not the β1 cytoplasmic domain, these results explain the selective interaction of this cytoplasmic domain with β3-endonexin. In addition, deletional analysis suggested that a core 91-residue sequence of β3-endonexin is sufficient for specific binding to the β3 cytoplasmic domain. These studies have identified a cytoplasmic domain sequence motif that specifies an integrin-specific protein-protein interaction.

where it is involved in the regulation of cell adhesion, migration, proliferation, and cell survival (2)(3)(4). The adhesive function of the integrin can be regulated by the cell (inside-out signaling), and the ligand-bound and clustered form of the integrin triggers cellular responses (outside-in signaling) (5).
Integrin signaling and the interaction of integrin cytoplasmic domains with intracellular signaling molecules are still poorly understood. For example, inside-out signaling is believed to involve interactions of integrin cytoplasmic domains with specific cytoplasmic elements. Studies of patients with rare defects in platelet aggregation or of recombinant human integrins expressed in various mammalian cells are consistent with a role for the ␤ 3 cytoplasmic domain in inside-out and outside-in signaling (6 -9). In addition, the over-expression of isolated ␤ cytoplasmic domains can disrupt or promote integrin signaling, conceivably by binding to factors that interact with the ␤ cytoplasmic domain (10 -12).
A few proteins have been found to interact directly with integrin cytoplasmic domains, and most of these studies have been performed in vitro (13). The ␣ IIb cytoplasmic domain has been reported to interact with calreticulin through a membrane-proximal GFFKR sequence that is highly conserved among all integrin ␣ subunits (14). Cytohesin-1 has recently been identified as a specific integrin ␤ 2 cytoplasmic domain binding protein (15), and direct interaction of filamin with this tail has been described (16). Other ␤ cytoplasmic domains have been found to interact with ␣-actinin, talin, pp125 FAK , and integrin-linked kinase (17)(18)(19)(20)(21). However, these latter interactions may not be specific for one particular ␤ cytoplasmic domain. Because most cells contain many different integrins, it is possible that cytoplasmic domain-specific binding proteins may exist that play a role in determining the specificity of integrin responses.
Recently, we identified a novel 111-amino acid polypeptide called ␤ 3 -endonexin, which is present in platelets, mononuclear lymphocytes, and several tissues, which interacts selectively with the ␤ 3 cytoplasmic domain in a yeast two-hybrid system (22). As a first step in assessing the potential biological functions of ␤ 3 -endonexin, we have conducted a systematic mutational analysis of both the ␤ 3 cytoplasmic domain and ␤ 3endonexin to identify amino acid residues required for this unique interaction.

MATERIALS AND METHODS
Antibodies-Monoclonal antibodies against the extracellular domain of the human ␤ 3 integrin subunit (monoclonal antibody 15) or the extracellular domain of the hamster ␤ 1 integrin subunit (monoclonal antibody 7E2) have been described (23,24). A monoclonal antibody against the extracellular domain of human ␤ 1 (antibody B-D15) was purchased from BioSource International (Camarillo, CA). A monoclonal antibody against the GAL4 DNA binding domain was purchased from Clontech Laboratories (Palo Alto, CA). * This work was supported by Grants HL 48728, AR 27214, and HL 56595 from the National Institutes of Health and by grants from Cor Therapeutics, Inc. This is publication 10305 from the Scripps Research Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Cell Culture and Transfection-CHO 1 cells stably expressing ␣ IIb ␤ 3 or ␣ IIb ␤ 3 ⌬717 (a ␤ 3 truncation mutant lacking the cytoplasmic domain) have been described (25). CHO cells stably expressing human integrin ␤ 1 paired with endogenous hamster ␣ subunits were obtained by transfecting ␤ 1 cDNA in pCDM8 using neomycin resistance as a selectable marker. Expression of human ␤ 3 or ␤ 1 integrins was quantified by Western blot technique using monoclonal antibodies 15 (5 g/ml) or B-D15 (1:400), respectively (22). The intensity of the bands of the ␤ subunits on scanned images of these Western blots was quantified by densitometry on a MacIntosh computer using NIH Image software (version 1.55). All labeled bands were analyzed within the linear range for the chemiluminescence reaction.
Binding of Integrin Cytoplasmic Domains to a ␤ 3 -Endonexin Affinity Matrix-Bacterial expression of ␤ 3 -endonexin as an amino-terminal histidine-tagged protein and preparation of a Nickel-agarose ␤ 3 -endonexin affinity resin were performed as described (22). Stably transfected CHO cell lines expressing approximately equivalent amounts of the indicated human integrins were lysed in 0.4 ml of lysis buffer containing 50 mM Tris, pH 7.2, 0.9% NaCl, 1 mM CaCl 2 , 1% Triton X-100, and protease inhibitors (100 units/ml aprotinin, 0.5 mM leupeptin, 4 mM Pefabloc, 0.1 mM E64) at 4°C for 30 min while shaking. Cell lysates were spun in a microfuge at 14,000 rpm for 20 min at 4°C. Then 0.35 ml of each supernatant were added to 0.35 ml of the lysis buffer containing no Triton, such that the final concentration of Triton was 0.5%. Each diluted lysate was batch-incubated with 1 ml of packed volume of ␤ 3 -endonexin affinity resin for 12 h at 4°C while shaking. Resins were then packed in columns and washed with 15 ml of lysis buffer, and bound proteins were eluted into 0.7-ml fractions of lysis buffer after the addition of 1 M imidazole. Fractions were collected and run on SDS-PAGE gels under nonreducing conditions. After electrotransfer to nitrocellulose, Western blotting was performed with monoclonal antibody 15 for ␤ 3 and B-D15 for ␤ 1 .
cDNA Constructs in Yeast Vectors-The yeast vector, pGBT9, was used to construct in frame fusions of integrin cytoplasmic domains (see Table I) with the GAL4 DNA binding domain (22). Truncations and point mutations in the carboxyl terminus of the ␤ 3 cytoplasmic domain were constructed by PCR using a common 5Ј primer (CGGAAGAGAG-TAGTAACAAAG) and a 3Ј primer encoding an appropriate stop codon or an amino acid change and a PstI restriction site. PCR products were cut with BamHI and PstI and cloned into BamHI-and PstI-cut pGBT9. cDNAs containing truncations in the amino terminus of the ␤ 3 cytoplasmic domain were constructed by PCR using a 5Ј primer encoding a BamHI restriction site and a 3Ј primer encoding the last seven residues of the ␤ 3 cytoplasmic domain, a stop codon and a PstI restriction site (GCTACTGCAGGTTAAGTGCCCCGGTACGTGATATTG). Resulting PCR products were cut with BamHI and PstI and ligated into pGBT9. Chimeras of the ␤ 3 and ␤ 1 cytoplasmic domains were constructed by splice overlap PCR mutagenesis and cloned into pGBT9 at the BamHI and PstI sites (26).
The yeast vector, pACT, was used to construct fusions of wild-type or truncated forms of ␤ 3 -endonexin cDNAs with the GAL4 activation domain (22,27). Amino-terminal truncations of ␤ 3 -endonexin were cloned into pACT by PCR using a 5Ј primer containing a BamHI restriction site and a common 3Ј primer (GATGCACAGTTGAAGTGAACTTGC). PCR products were cut with BamHI and XhoI and cloned into pACT. Carboxyl-terminal truncated forms of ␤ 3 -endonexin were constructed by splice overlap PCR, and the PCR products were cut with BamHI and XhoI and ligated into BamHI-and XhoI-cut pACT.
Yeast Two-hybrid System-Yeast strain maintenance and transformation have been described (22). The yeast two-hybrid system was used to quantify the extent of binary interactions between ␤ 3 -endonexin and integrin ␤ cytoplasmic domains. Protein expression in transformed yeast was analyzed by SDS-PAGE and Western blotting using a specific monoclonal antibody for the GAL4 DNA binding domain (28). The extent of expression of the reporter gene, lacZ, was determined by quantitative liquid ␤-galactosidase assay (29) and taken as an indicator of the strength of interaction between the two fusion proteins (29 -32). A one-tailed Student's t test for unpaired samples was used for statistical calculations.

Interaction of the ␤ 3 Integrin Subunit with ␤ 3 -Endonexin
Requires the Integrin Cytoplasmic Domain-Previous studies have shown that ␤ 3 -endonexin binds to the cytoplasmic domain of the ␤ 3 integrin subunit when the isolated cytoplasmic domain is expressed in a yeast two-hybrid system (22). This interaction is structurally specific, because it was not observed with the cytoplasmic domains of the integrin ␣ IIb , ␤ 1 , or ␤ 2 subunits. Therefore, we conducted a mutational analysis using the yeast two-hybrid system to identify sites within these two proteins that are necessary for this binary interaction. Prior to undertaking such an analysis, experiments were performed to assess the specificity with which ␤ 3 -endonexin binds to the ␤ 3 cytoplasmic domain in the context of an intact integrin.
CHO cell lines were prepared that stably expressed the human ␣ IIb subunit paired with either human ␤ 3 or ␤ 3 ⌬717, a truncated form of ␤ 3 missing the entire cytoplasmic domain except for a putative membrane-proximal lysine residue (␤ 3 Lys 716 ) (33). As an additional control, a CHO cell line expressing human ␤ 1 paired with endogenous hamster ␣ subunits was prepared. Analyzing these cell lines by SDS-PAGE and Western blotting using antibodies specific for the extracellular portion of human ␤ 3 or ␤ 1 showed that they expressed similar levels of their respective ␤ subunits (data not shown). As a source of ␤ 3 -endonexin for in vitro binding studies, a histidinetagged form of ␤ 3 -endonexin was expressed in bacteria and coupled noncovalently to a metal chelation affinity resin. Then 1 The abbreviations used are: CHO, Chinese hamster ovary; PAGE, polyacrylamide gel electrophoresis; PCR, polymerase chain reaction.
FIG. 1. The cytoplasmic domain of the integrin ␤ 3 subunit is required for interaction with ␤ 3 -endonexin. As described under "Materials and Methods," an affinity matrix was prepared with histidine-tagged ␤ 3 -endonexin bound noncovalently to a metal chelation resin. CHO cell lysates (0.7 ml) containing the full-length ␤ 3 integrin subunit (A), the ␤ 3 ⌬717 subunit (B), or the full-length ␤ 1 subunit (C) were incubated with the affinity resin for 12 h at 4°C. After washing, proteins were eluted from the resin in 0.7 ml of lysis buffer containing 1 M imidazole. Lysate (lanes 1), last column wash (lanes 2), and resin eluate (lanes 3) were then analyzed by SDS-PAGE under nonreducing conditions, transferred to nitrocellulose, and immunoblotted with monoclonal antibodies specific for the extracellular domain of ␤ 3 (A and B) or ␤ 1 (C). ␤ 1 integrins appear as a double band representing ␤ 1 and a precursor form of this integrin subunit. The autoradiographs shown are representative for three separate experiments. Integrin bands on the Western blots were analyzed by densitometry. Integrins in the elution fraction were expressed as percentages of integrins in the lysate (which was defined as 100%). The data represent the means Ϯ S.E. of three separate experiments. equal aliquots of affinity resin were incubated with equal volumes of detergent extracts from each of the three CHO cell lines, and the amount of human integrin ␤ subunit retained on the ␤ 3 -endonexin resin was determined by Western blotting (Fig. 1). Approximately 53% of the full-length ␤ 3 integrin subunit that was applied to the ␤ 3 -endonexin affinity matrix was retained, compared with only 14% of ␤ 3 ⌬717 and 5% of the ␤ 1 integrin subunit. The differences between ␤ 3 and ␤ 3 ⌬717 and between ␤ 3 and ␤ 1 were significant (p Ͻ 0.006). In contrast, the difference between ␤ 3 ⌬717 and ␤ 1 was not (p Ͼ 0.05) (Fig. 1). Furthermore, using a monoclonal antibody specific for hamster ␤ 1 (monoclonal antibody 7E2) to detect ␤ 1 in CHO cell lysates, no significant binding of endogenous hamster ␤ 1 integrin to ␤ 3 -endonexin could be detected by Western blotting (data not shown). These results demonstrate that interaction of the fulllength ␤ 3 integrin subunit with ␤ 3 -endonexin is mediated by the integrin cytoplasmic domain.
Both Membrane-proximal and Membrane-distal Residues of the ␤ 3 Cytoplasmic Domain Are Necessary for Interaction with ␤ 3 -Endonexin-To study the structural basis for the interaction between the ␤ 3 cytoplasmic domain and ␤ 3 -endonexin in more detail, a series of truncation mutants of the ␤ 3 cytoplasmic domain (Table I) were fused in-frame to the carboxyl terminus of the GAL4 DNA binding domain and co-expressed in yeast with ␤ 3 -endonexin fused to the GAL4 DNA activation domain. In this system, the extent of expression of the reporter gene, lacZ, can be taken as an indication of the strength of interaction between the two fusion proteins (29 -32). Deletion of the carboxyl-terminal 1-3 amino acids from the cytoplasmic domain of ␤ 3 (␤ 3 ⌬762, ␤ 3 ⌬761, or ␤ 3 ⌬760) caused no significant reduction in its interaction with ␤ 3 -endonexin (Fig. 2). In fact, deletion of the carboxyl-terminal threonine residue actually increased apparent binding (p Ͻ 0.0002) (Fig. 2). However, deletion of the carboxyl-terminal 4 residues from the ␤ 3 cytoplasmic domain (␤ 3 ⌬759) reduced binding to ␤ 3 -endonexin, whereas deletion of 8 residues from the carboxyl terminus of the ␤ 3 cytoplasmic domain (␤ 3 ⌬755) virtually abolished binding (Fig. 2). This result was not due to lack of expression of any of the GAL4 DNA binding domain fusion proteins (Fig. 3).
To further elucidate the role of the carboxyl terminus of the ␤ 3 cytoplasmic domain for interaction with ␤ 3 -endonexin, we attached the carboxyl-terminal 7 residues of ␤ 3 to the ␣ IIb cytoplasmic domain (␣ IIb 989 -1008/␤ 3 756 -762) ( Table I). Although expressed in yeast, this construct did not interact with ␤ 3 -endonexin (Figs. 2 and 3), indicating that the carboxylterminal region of the integrin ␤ 3 cytoplasmic domain is necessary but not sufficient for the interaction with ␤ 3 -endonexin.
To determine whether residues in the membrane-proximal region of the ␤ 3 cytoplasmic domain are necessary for the interaction with ␤ 3 -endonexin, the effects of amino-terminal truncations of the ␤ 3 cytoplasmic domain were assessed. Deletion of even a single residue from the amino terminus (Lys 716 ) caused a more than 92% reduction in binding to ␤ 3 -endonexin (p Ͻ 0.0001). Additional constructs containing deletions of 3, 6, or 11 residues from the amino terminus of the ␤ 3 cytoplasmic domain also failed to interact (data not shown).
Taken together with the results of the carboxyl-terminal cytoplasmic domain truncations, these data indicate that both the amino and carboxyl termini of the ␤ 3 cytoplasmic domain are required for the interaction with ␤ 3 -endonexin. This could mean that both membrane-proximal and membrane-distal regions of the ␤ 3 cytoplasmic domain are directly involved in binding and/or that overall folding of the cytoplasmic domain is a critical determinant of its interaction with ␤ 3 -endonexin.
The NITY Motif at ␤ 3 756 -759 Is Critical for Interaction of the ␤ 3 Cytoplasmic Domain with ␤ 3 -Endonexin-Despite the fact that the ␤ 3 and ␤ 1 cytoplasmic domains exhibit high overall similarity (60% identical; 68% identical plus conservative substitutions) only the ␤ 3 cytoplasmic domain interacts with ␤ 3 -endonexin ( Fig. 1) (22). It is therefore of particular interest to identify the residues within the ␤ 3 cytoplasmic domain that account for the specific binding to ␤ 3 -endonexin. Because the region of greatest dissimilarity between these two cytoplasmic domains is at the extreme carboxyl terminus (Table I)  KVGFFKRNRPPLEEDDEEGQ NITY RGT influence the ability of these domains to interact with ␤ 3endonexin. Indeed, exchange of carboxyl-terminal 7 residues of the ␤ 3 cytoplasmic domain amino acids 756 -762) for the corresponding region of the ␤ 1 cytoplasmic domain resulted in a strong interaction of the new chimeric ␤ 1 /␤ 3 cytoplasmic domain with ␤ 3 -endonexin (Fig. 4). In fact, exchange of only 4 ␤ 3 residues 756 -759 (NITY) for the corresponding residues in ␤ 1 (NPKY) or exchange of even a single amino acid of ␤ 3 (Ile 757 ) for Pro 773 in ␤ 1 (␤ 1 P773I) now endowed the ␤ 1 cytoplasmic domain with the ability to interact with ␤ 3 -endonexin (Fig. 4). Conversely, swapping the carboxyl-terminal 7 residues of the ␤ 1 cytoplasmic domain into the corresponding region of ␤ 3 or exchange of the NPKY motif at ␤ 1 772-775 for the NITY motif in ␤ 3 abolished interaction of these chimeric cytoplasmic domains with ␤ 3 -endonexin. Moreover, introduction of Pro 773 of ␤ 1 into the ␤ 3 cytoplasmic domain, resulting in ␤ 3 I757P, decreased binding to ␤ 3 -endonexin by more than 70% (p Ͻ 0.004) (Fig. 4). These data could not be accounted for by differences in levels of expression of the ␤ 3 I757P and ␤ 1 P773I fusion proteins (Fig. 3). These data indicate that the ␤ 3 linear sequence, 756 NITY, is critical for the interaction of the ␤ 3 cytoplasmic domain with ␤ 3 -endonexin.
To examine the importance of individual components of the NITY motif further, alanine was substituted individually for each amino acid in this motif. Alanine substitution at Ile 757 (␤ 3 I757A) or Tyr 759 (␤ 3 Y759A) in ␤ 3 resulted in a 75 or 92% reduction in interaction of the ␤ 3 cytoplasmic domain with ␤ 3 -endonexin, respectively (Fig. 5). On the other hand, more conservative substitutions of Ile 757 (␤ 3 I757L) or Tyr 759 (␤ 3 Y759F) or alanine substitutions of Asn 756 (␤ 3 N756A) or Thr 758 (␤ 3 T758A) had little or no effect on binding (Fig. 5). This demonstrates that Ile 757 and Tyr 759 in ␤ 3 are critical residues for interaction with ␤ 3 -endonexin. As shown in Fig. 6, despite its dissimilarity with the corresponding region in several other human ␤ cytoplasmic domains, the NITY motif is highly conserved among ␤ 3 integrins of various species. Thus, we propose that this motif is responsible for the ␤ subunit specificity of ␤ 3 -endonexin.
Given the key role of the NITY motif in this interaction, it should be noted that this motif is also important for localization of ␤ 3 integrins to focal adhesions and for integrin signaling (34). For example, deletion of ␤ 3 residues 759 YRGT (␤ 3 ⌬759) significantly reduced cell spreading and recruitment of ␤ 3 integrins to focal adhesion sites, whereas deletion of ␤ 3 residues 757 ITYRGT (␤ 3 ⌬757) totally abolished cell spreading and formation of focal contacts (34). Retaining Ile 757 partially pre- served these functions. The point mutation ␤ 3 Y759A also significantly reduced cell spreading and ␤ 3 integrin recruitment to focal adhesions (34). Thus, the region of the ␤ 3 cytoplasmic domain that is necessary for interaction with ␤ 3 -endonexin also is necessary for post-adhesive functions of ␤ 3 integrins. Whether ␤ 3 -endonexin modulates the adhesive or post-adhesive functions of ␤ 3 integrins remains to be determined. In this context, preliminary studies show that overexpression of ␤ 3 -endonexin in an ␣ IIb ␤ 3 /CHO cell model system increases the affinity state of integrin ␣ IIb ␤ 3 (35).
Recent studies have demonstrated tyrosine phosphorylation of the integrin ␤ 3 cytoplasmic domain as well as calpain-induced cleavage at various sites within the ␤ 3 cytoplasmic domain during thrombin-induced activation of platelets (36,37). Although the latter might be expected to cause release of ␤ 3endonexin from the ␤ 3 cytoplasmic domain, we cannot predict the effects of tyrosine phosphorylation on the binding of ␤ 3endonexin, and we have not observed tyrosine phosphorylation of the ␤ 3 cytoplasmic domain in the yeast system. 2 Both Amino-and Carboxyl-terminal Residues of ␤ 3 -Endonexin Are Required for Interaction with the ␤ 3 Cytoplasmic Domain-As a first approach to identify the residues within ␤ 3endonexin that are critical for binding to the ␤ 3 cytoplasmic domain, we investigated the binding of carboxyl-terminal and amino-terminal truncation mutants of ␤ 3 -endonexin to this cytoplasmic domain. The carboxyl terminus of ␤ 3 -endonexin contains three heptad repeats that may form coiled-coil structures (residues 89 -111) (38). Removal of part of the last of these repeats by deleting 2 amino acids from the carboxyl terminus of ␤ 3 -endonexin (residues 110 and 111 of ␤ 3 -endonexin, EN⌬2) decreased binding to the ␤ 3 cytoplasmic domain more than 80% (Fig. 7). Deletion of 7 or more amino acids (corresponding to at least one heptad repeat) from the carboxyl terminus of ␤ 3 -endonexin abolished binding completely (Fig. 7). Furthermore, deletion of 9 or 20 residues from the amino ter-minus of ␤ 3 -endonexin had no effect on binding, whereas deletion of 35 or more amino-terminal residues abolished binding (Fig. 7). These results show that a construct containing only ␤ 3 -endonexin residues 21-111 was sufficient for the interaction with the ␤ 3 cytoplasmic domain (Fig. 7). Thus, the amino terminus of ␤ 3 -endonexin is dispensible for interaction of this polypeptide with the ␤ 3 cytoplasmic domain, but the carboxyl terminus is not.
In addition to determining the basis for the selectivity of ␤ 3 -endonexin for the ␤ 3 cytoplasmic domain, the present results will prove useful in designing studies aimed at establishing the physiological function of ␤ 3 -endonexin. For example, the established importance of the NITY motif in the ␤ 3 cytoplasmic domain in certain aspects of integrin signaling and in binding of ␤ 3 -endonexin suggests that over-expression of ␤ 3endonexin or incorporation of NITY-containing peptides into cells might disrupt (or promote) ␤ 3 integrin functions such as ligand binding, cell spreading, or modulation of gene expression (39). These effects could be specific for ␤ 3 integrins. If so, this would support the idea that the specificity of cellular responses to integrin ligands is determined by several factors, including the composition of the extracellular matrix, the integrin repertoire of the cell, and the intracellular complement and function of integrin cytoplasmic domain-binding proteins.