Molecular requirements for assembly and function of a minimized human integrin alphaIIbbeta3.

Integrin subunit compatibility within and between species plays a major role in heterodimer assembly and ligand specificity. As an example, human αIIb pairs only with human β3 and does not assemble a heterodimer with β3 from other species. We use interspecies subunit chimeras to identify molecular requirements for subunit compatibility and show that species-restricted heterodimer assembly depends on a unique hexapeptide VGSDNH in an extended loop of the hypothetical human β3 MIDAS domain. This allows us to express αIIb(1-233) and β3(111-318) as a soluble, mini-integrin that retains RGD-dependent ligand recognition. Thus, in the case of one integrin, αIIbβ3, the molecular requirements for integrin subunit compatibility and ligand recognition are intimately related.

Integrin subunit compatibility within and between species plays a major role in heterodimer assembly and ligand specificity. As an example, human ␣ IIb pairs only with human ␤ 3 and does not assemble a heterodimer with ␤ 3 from other species. We use interspecies subunit chimeras to identify molecular requirements for subunit compatibility and show that species-restricted heterodimer assembly depends on a unique hexapeptide VGSDNH in an extended loop of the hypothetical human ␤ 3 MIDAS domain. This allows us to express ␣ IIb (1-233) and ␤ 3 (111-318) as a soluble, mini-integrin that retains RGD-dependent ligand recognition. Thus, in the case of one integrin, ␣ IIb ␤ 3 , the molecular requirements for integrin subunit compatibility and ligand recognition are intimately related.
␣␤ subunit compatibility is fundamentally important to integrin biology since it determines the success of heterodimer assembly and is a predominant factor influencing receptor specificity (1,2). In this regard, human ␣ IIb is highly restrictive, pairing exclusively with human ␤ 3 (3). Heterodimer assembly between analogous subunits from different species is subject to the same compatibility restrictions. For example, human ␣ IIb does not form a functional heterodimer with ␤ 3 subunits from other species, including Xenopus and avian species.
While the expression of membrane-associated ␣ IIb ␤ 3 depends upon intracellular subunit processing or trafficking that is regulated to a certain extent by cytoplasmic sequences (4), heterodimer assembly and function are retained by soluble subunits composed entirely of extracellular domains (5,6). Moreover, soluble heterodimers retain the same subunit compatibility restrictions characteristic of the full-sized integrin. Thus, the molecular cues for both selective heterodimer formation and function must reside within extracellular sequences peculiar to each integrin subunit.
In this study, we sought to ascertain the sequences of each subunit that are required for heterodimer assembly in the case of the well characterized integrin ␣ IIb ␤ 3 .

EXPERIMENTAL PROCEDURES
Synthesis of Interspecies ␤ 3 Subunit Chimeras-Extracellular domains of human ␣ IIb and ␤ 3 were expressed by co-infection of Trichoplusia ni (High Five) insect cells with recombinant baculoviruses (7)(8)(9). Soluble human ␣ IIb (residues 1-964) was produced from a 3-kilobase cDNA fragment encoding the extracellular domains, and the signal sequence was isolated as an EcoRI fragment from a Bluescript clone (gift of Dr. David R. Phillips, COR Therapeutics, Inc., South San Francisco) and ligated into pVL1392 (Invitrogen). Sequence accuracy of recombinant ␣ IIb (clone IIb.5) was confirmed using Sequenase 2.0 (U. S. Biochemical Corp.). ␤ 3 clones represent residues 1-690, with respect to the numbering of the mature human sequence. Chimera 2.10 was produced from 2.9 (9) and inserted into pVL1392. To create 2.11, 2.12, and 2.13, an EcoRI-SacI 812-base pair fragment of 2.9 was reconstructed by splice-overlap extension polymerase chain reaction. For each construct, primers were designed so as to exchange the avian coding sequence with the analogous human sequence (sites A and B). cDNA constructs were cloned into pVL1392 or pVL1393, purity of recombinant viral clones was confirmed by polymerase chain reaction, and sequence accuracy was established using Sequenase 2.0. Individual integrin subunits and heterodimeric complexes were expressed in T. ni High Five cells (Invitrogen) grown to 1 ϫ 10 6 cells/ml in serum-free EXCell 400 medium (JRH Biosciences, Lenexa, KS). Cells were coinfected in culture flasks with clone IIb.5 at a multiplicity of infection of 15 and one of the ␤ 3 chimeras (multiplicity of infection ϭ 1.5). Culture media were harvested 4 days after infection, cellular debris was removed by centrifugation, and protease inhibitors were added (1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 0.01 mM leupeptin, 1 mM pepstatin A, and 5 mM N-ethylmaleimide) prior to storage at 4°C for up to 2 weeks.
Expression of Functional Integrin Heterodimers-Assembly of the human ␣ IIb subunit into ␣ IIb ␤ 3 heterodimers requires that specific human sequences be represented within the ␤ 3 subunit. The ␣ IIb subunit is secreted at concentrations ranging from 0.05 to 0.2 g/ml, and each of the ␤ 3 subunits is secreted in (several)-fold excess of ␣ IIb , i.e. at concentrations ranging from 0.5 to 1.0 g/ml. When co-infection results in secretion of heterodimers, the concentration of these typically ranges from 0.1 to 0.2 g/ml.
Aliquots of media from cells infected with recombinant viruses encoding integrin proteins were subjected to native acrylamide gel electrophoresis in the absence of SDS to separate ␣ IIb ␤ 3 heterodimers from other proteins. Protein samples were added in 50 mM Tris-HCl, 15% glycerol (v/v), 0.01% bromphenol blue, 2 mM CaCl 2 , pH 6.8, to polyacrylamide slab gels (5% resolving; 3% stacking) cast without SDS. Cast gels and electrophoresis chambers were chilled to 4°C for at least 4 h prior to use. The running buffer (20 mM Tris-HCl, 192 mM glycine, 2 mM CaCl 2 , pH 8.3, at room temperature) was cooled to 4°C. Protein samples were loaded (20 l per lane), and electrophoresis was conducted for 3-5 h at 10 mA/3 mm of gel, maintaining the gel chamber and buffer at 4°C. Resolved proteins were transferred to polyvinyl pyrolidine filters (PVDF) 1 by electrophoresis in fresh buffer at 4°C, maintaining a constant voltage (30 V) for 18 h. Unreacted PVDF sites were blocked by incubating the membrane in 10% nonfat milk for 30 min at ambient temperature with gentle agitation. The membrane was then incubated  1 The abbreviation used is: PVDF, polyvinyl pyrolidine filters. in 10 ml of 5 g/ml AP2 or OPG2, diluted in 50 mM Tris, 150 mM NaCl, 2 mM CaCl 2 , 0.1% (v/v) Tween, pH 7.4 (TBST), for 2 h at 22°C. These murine monoclonal IgG antibodies, AP2 and OPG2, bind epitopes expressed only in the context of the complex and serve to detect heterodimers. Moreover, OPG2 is an RGD mimetic that contains the RYD sequence in an extended loop of its antigen binding site (10,11). Individual ␣ IIb or ␤ 3 subunits and the individual fragments derived from them, regardless of species origin, are never bound by AP2, OPG2, or the RGDW matrix. Unbound primary antibody was removed by three successive rinses with TBST, and membranes were exposed to an excess of horseradish peroxidase-conjugated donkey anti-murine IgG (1:5000 dilution; Jackson Immunologicals, Westgrove, PA). Following 1 h at 22°C, unbound secondary antibody was removed by three TBST rinses, and bound antibody was detected by the ECL method (Amersham) according to the manufacturer's directions. No protein bands were visualized using monoclonal antibody 12F1, specific for the integrin ␣ 2 ␤ 1 (hybridoma cell line is a generous gift from Dr. V. Woods, University of California, San Diego, La Jolla, CA), or AP1, specific for glycoprotein Ib heavy chain (7) (not shown). Affinity Purification of Heterodimers by RGDW-Sepharose Chromatography-Culture media from cells co-infected with viruses containing human ␣ IIb and ␤ 3 subunits were incubated for 24 h at 4°C with RGDW-coupled Sepharose 4B beads (Pharmacia, Uppsala, Sweden). Buffer was then aspirated, and packed beads were thrice rinsed in 20 volumes of TBST plus 2 mM CaCl 2 with agitation. Aliquots of each rinse solution were saved for analysis, and bound heterodimers were eluted by three successive rinses in equal volumes of 3 mM RGDW peptide in TBST plus 2 mM CaCl 2 .
Proteins in each fraction were separated by standard (denaturing) SDS-polyacrylamide gel electrophoresis in preparation for detection of integrin subunits. Aliquots of each fraction were mixed with SDSsample buffer to yield the following final concentrations of each component: 50 mM Tris-HCl, 15% glycerol (v/v), 1% SDS (w/v), 5% 2-mercaptoethanol (v/v), and 0.01% bromphenol blue, pH 6.8. Mixtures were covered and incubated in a boiling water bath for 10 min. 20 l of each mixture were applied to a 7.5% polyacrylamide gel using a 3% stacking gel. The electrophoresis buffer was 192 mM glycine, 20 mM Tris, 0.1% (v/v) SDS, pH 8.3. Separated proteins were transferred to PVDF membranes, and the ␣ IIb or ␤ 3 subunit in each aliquot was detected by the binding of rabbit polyclonal anti-human ␣ IIb plus ␤ 3 IgG (10 g/ml in TBST plus 10 mM EDTA). Bound rabbit IgG was visualized by addition of horseradish peroxidase-conjugated goat anti-rabbit IgG (1:1000 dilution, Jackson Immunologicals) using the ECL method as described above for murine monoclonal IgG.
Expression of the Mini-integrin-The preceding methods were employed for the synthesis, purification, and detection of the mini-integrin with the following exception. Following RGDW-Sepharose affinity chromatography, proteins in fractions were separated by native gel electrophoresis using a 6% acrylamide separating gel cast in the presence of 2 mM CaCl 2 and 1.5% (v/v) glycerol. The stacking gel (2.5% acrylamide) was cast in the presence of 2 mM CaCl 2 and 3% (v/v) glycerol. Acrylamide gels and apparati were chilled to 4°C. Electrophoresis was conducted for 5 h at 25 mAmp per 3 mm maintaining a fixed temperature of 4°C. Separated proteins were transferred to PVDF membranes (Immobilon-P SQ ; Bedford, MA) using constant voltage (85 V) for 1 h at 4°C.

RESULTS AND DISCUSSION
A comparison of interspecies ␤ 3 chimeric subunits (Fig. 1) identifies a restricted human sequence that is required for heterodimer assembly with human ␣ IIb . In Western blots of proteins separated by native polyacrylamide gel electrophoresis (in the absence of SDS), only ␤ 3 subunits containing human sequence 133-323 form heterodimers (Fig. 2). For example, the ␤ 3 subunit 2.3, which contains predominantly human sequence, but avian sequence within residues 133-323, is not incorporated into a heterodimer, while the inverse construct, ␤ 3 subunit 2.10, containing human sequence as residues 133-323 within an avian background, is isolated in heterodimers that are detected by either AP2 or OPG2 (Fig. 2) and bind to RGDW-Sepharose. Regardless of the extent to which the ␤ 3 subunit is incorporated into heterodimers, equivalent amounts of ␣ IIb and ␤ 3 molecules are secreted, based on the binding of PMI-1, specific for ␣ IIb 844 -860 (12), and AP5 or AP6, specific for ␤ 3 1-6 and ␤ 3 214 -219, respectively (7, 13) (not shown).
Thus, the human ␤ 3 sequence necessary for subunit compatibility is confined to residues 133-323 within the putative ␤ 3 MIDAS domain (14). A comparison of this sequence between diverse species (Fig. 3) reveals isolated single amino acid differences and, more strikingly, two relatively long stretches of nonhomologous sequence that we designate site A (residues 179 -183) and site B (residues 275-280). This finding compelled us to test the relevance of each of these sequences to subunit compatibility.
␤ 3 subunit 2.11, containing human site A alone, does not associate with human ␣ IIb (Fig. 2). On the other hand, ␤ 3 subunit 2.12, containing human site B alone, forms a complex with the human ␣ IIb that is detected by AP2 or OPG2 (Fig. 2). ␤ 3 subunit 2.13, containing both human sites A and B, behaves exactly as does ␤ 3 subunit 2.12. Thus, the hexapeptide sequence VGSDNH (site B of human ␤ 3 ) is necessary for speciesrelated association of the ␤ 3 subunit with human ␣ IIb . In every instance, heterodimer assembly results in a receptor that is capable of binding specifically to the RGDW matrix. The results with the human ␣ IIb ⅐human ␤ 3 heterodimer are depicted in Fig. 4, but identical findings were made with human ␣ IIb coexpressed with each of the chimeric ␤ 3 subunits 2.10, 2.12, or 2.13.
The ␤ 3 MIDAS domain is structurally analogous to integrin ␣ subunit I domains from ␣ L and ␣ M , which can be expressed as isolated entities (15)(16)(17), so it is not surprising that it too can be expressed as an isolated domain, as we describe below. However, our parallel objective was to identify a short segment of the ␣ IIb subunit that is also necessary for heterodimer assem- bly. Our strategy drew upon our observations with the ␤ 3 subunit that sequences required for function and heterodimer assembly are intimately associated if not identical. Thus, we focused first on those regions of ␣ IIb considered to be important to receptor function. Ligand binding specificity of ␣ IIb has recently been shown to lie within the amino one-third of that subunit (18), which includes the first two of four divalent cation binding sites (19). At the same time, deletion of any one of these four divalent cation binding sites has no effect on heterodimer formation in COS cells transfected with full-length, membraneassociated subunits (20). Thus, we elected to test a segment of ␣ IIb representing the amino-terminal domain up to but not including the first divalent cation binding site, namely, residues 1-233.
This strategy was successful. High Five cells co-infected with viruses encoding ␣ IIb 1-233 and ␤ 3 111-318 secrete heterodimers formed by these short segments that can be isolated from culture media by RGDW-Sepharose affinity chromatography and then detected in eluate fractions by the binding of OPG2 (Fig. 5) or AP2 (not shown). Thus, these limited subunit segments retain sufficient structural integrity to facilitate as-sembly of a functional receptor. This represents the strongest proof of the importance of these domains to subunit compatibility and receptor functions.
The ␤ 3 MIDAS domain (14) contains several oxygenated residues that are proposed to form a divalent cation binding coordination site. The proper coordination of cation within this domain is crucial to ligand binding (14,21). Residues 275-280 of human ␤ 3 would exist within an extended loop of this domain, based upon any algorithm for the prediction of secondary structure (22,23). By direct comparisons with the I domain crystal structures, we speculate that ␤ 3 275-280 would likely correspond to the ␤E-␣6 loop of integrin ␣ M (24), which is analogous to the ␤4-␣6 loop of the ␣ L I domain (25). In either case, this extended loop is not predicted to participate directly in coordination of the cation, but it is required for the maintenance of structures that stabilize an active conformation of the ␣ L I domain (25). Direct confirmation of our sequence assignments must, of course, await crystallization of this region of the ␤ 3 subunit.
Since human ␣ IIb selectively associates only with human ␤ 3 , there must be sequence and/or structural cues peculiar to the 275-280 loop of human ␤ 3 . From a comparison of the analogous sequences in the remaining seven human ␤ subunits, it is striking that ␤ 3 275-280 remains unique (Fig. 6). We postulate that a complementary site of ␣ IIb located within residues 1-233 FIG. 3. Identification of a unique human ␤ 3 integrin sequence required for heterodimer formation. The human ␤ 3 sequence 129 -323 (26), encompassing the MIDAS domain (14,25), is aligned with the analogous sequences of rat (27), mouse (27), chicken (avian) (28), and Xenopus (29) ␤ 3 . Horizontal bars represent sequence identity. ␤ 3 sequences corresponding to defined structures (␣helices or ␤-strands) of the integrin ␣ L subunit I domain (25) are indicated. Extended loop regions designated A and B, with high nonidentity between species, were further investigated as potential determinants of species-restricted heterodimer formation. interacts specifically with this flexible peptide segment and that this association influences the overall tertiary structure of both the ␣ IIb and the ␤ 3 subunits, thereby permitting heterodimer assembly to occur. While it is tempting to speculate that comparable sequences regulate subunit compatibility and heterodimer assembly of all integrins, in general, we are compelled to point out that our findings have been made only with one integrin, ␣ IIb ␤ 3 , and may thus reflect properties unique to this integrin.
We show that molecular compatibility between otherwise complex polypeptides like ␣ IIb or ␤ 3 can depend upon a short sequence, such as VGSDNH in ␤ 3 . It is not coincidental that this key sequence represents an extended loop of the ␤ 3 MIDAS domain, thought to be involved in divalent cation coordination in a manner critical to receptor function. There must be an intimate relationship between structural cues that regulate heterodimer assembly and ligand recognition. We propose that during receptor assembly, the initial contact between key sequences of each subunit initiates conformational changes in the opposite subunit resulting in the formation of an effective ligand recognition pocket. Thus, these key sequences on each subunit initially occupy at least a portion of the recognition pocket. Our findings validate the proposal (3) that subunit interactions, influenced by cation coordination, would then be disrupted or displaced by ligand binding.