Transmembrane Domain Helix Packing Stabilizes Integrin αIIbβ3 in the Low Affinity State*

Regulated changes in the affinity of integrin adhesion receptors (“activation”) play an important role in numerous biological functions including hemostasis, the immune response, and cell migration. Physiological integrin activation is the result of conformational changes in the extracellular domain initiated by the binding of cytoplasmic proteins to integrin cytoplasmic domains. The conformational changes in the extracellular domain are likely caused by disruption of intersubunit interactions between the α and β transmembrane (TM) and cytoplasmic domains. Here, we reasoned that mutation of residues contributing to α/β interactions that stabilize the low affinity state should lead to integrin activation. Thus, we subjected the entire intracellular domain of the β3 integrin subunit to unbiased random mutagenesis and selected it for activated mutants. 25 unique activating mutations were identified in the TM and membrane-proximal cytoplasmic domain. In contrast, no activating mutations were identified in the more distal cytoplasmic tail, suggesting that this region is dispensable for the maintenance of the inactive state. Among the 13 novel TM domain mutations that lead to integrin activation were several informative point mutations that, in combination with computational modeling, suggested the existence of a specific TM helix-helix packing interface that maintains the low affinity state. The interactions predicted by the model were used to identify additional activating mutations in both the α and β TM domains. Therefore, we propose that helical packing of the α and β TM domains forms a clasp that regulates integrin activation.

Integrin heterodimers are essential for the development and functioning of multicellular animals, because they mediate cell migration and cell adhesion and can influence gene expression and cell proliferation (1). All of the integrin heterodimers are composed of single pass Type I transmembrane (TM) 1 protein subunits ␣ and ␤. A central feature of these receptors is their capacity for rapid changes in their adhesive function mediated by changes in their ligand binding affinity, operationally defined here as "activation." The prototypical integrin, platelet ␣IIb␤3, is activated through interactions of the cytoplasmic integrin tails (ϳ20 and 47 residues for ␣ and ␤ tails, respectively) with intracellular proteins such as talin (2). These interactions initiate a long-range conformational change in the large extracellular domains (Ͼ700 residues each), resulting in high affinity binding of fibrinogen, von Willebrand factor, and fibronectin and consequently platelet aggregation and adherence to the vessel wall (1).
Initial mutational studies suggested that a salt bridge between ␣IIbArg 995 and ␤3Asp 723 helps maintain the integrin in the low affinity state by forming part of an interactive face between ␣ and ␤ subunit cytoplasmic domains (3). Protein engineering studies from Springer laboratory have further advanced the idea that specific integrin ␣/␤ interactions maintain the low affinity conformation of the receptor. In particular, enforced association of either the C-terminal region of the extracellular domains (4) or that of the cytoplasmic domains (5) leads to expression of an inactive integrin. Furthermore, during physiological integrin activation, changes in fluorescence resonance energy transfer between fluorophores joined to the ␣L and ␤2 cytoplasmic domains are consistent with cytoplasmic domain separation (6). Finally, constraining the integrin ␣ and ␤ transmembrane domains with intersubunit disulfide bonds blocks integrin activation from inside the cell. However, this constraint does not prevent activation by divalent cations and antibodies that activate by binding to the extracellular domain (7). Taken together, these data suggest that default low affinity state of integrins is maintained by interactions between integrin ␣ and ␤ subunits and that physiological activation occurs when cytoplasmic domain ligands, such as talin, disrupt these interactions. Support for the idea that an Arg 995 -Asp 723 salt bridge is an important constraint for the low affinity state comes from a NMR spectroscopy study (8). Specifically, in isolated ␣IIb and ␤3 cytoplasmic domain peptides, the salt bridge was identified as part of a helical interface between the membrane-proximal regions of ␣ and ␤ subunits. Furthermore, this interaction was disrupted by talin, supporting the notion that disruption of this salt bridge is involved in integrin activation (8). However, other groups have failed to observe intersubunit interactions in the membrane-proximal region, suggesting that it is of relatively low affinity (9,10). Therefore, additional intracellular regions of the receptor could contribute ␣/␤ interactions to "clasp" it into the low affinity state. Indeed, in vitro model systems identified heterodimeric interactions between integrin ␣ and ␤ TM domains (11) and such interactions have also been suggested by molecular modeling (12,13) and disulfide crosslinking approaches (7). Mutation of residues contributing to ␣/␤ interactions that stabilize the low affinity state should lead to integrin activation. Thus, we subjected the entire intracellular domain (cytoplasmic plus TM domain in Fig. 1) of the ␤3 integrin subunit to unbiased random mutagenesis and selected it for activated mutants. Through this analysis, we have confirmed the importance of the membrane-proximal domain in maintenance of the low affinity state. In contrast, no activating mutations were identified in the more distal cytoplasmic tail suggesting that this region is dispensable for the maintenance of the inactive state. This approach also identified 13 novel TM domain mutations that lead to integrin activation. Among these were several informative point mutations that suggested the existence of a TM helix-packing interface that maintains the low affinity state. Computational modeling indicates that these mutations disrupted intersubunit interactions either directly or indirectly by altering helical length/tilt angle. The interactions predicted in the model were used to create additional activating mutations in both the ␣ and ␤ TM domains. Therefore, we propose that ␣ and ␤ TM regions interact to form a clasp that constrains integrin activation.

EXPERIMENTAL PROCEDURES
Cell Culture, Cell Lines, and Reagents-Chinese hamster ovary (CHO) cells were obtained from American Type Culture Collection (ATCC) and cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 1% non-essential amino acids (Sigma), 50 units of penicillin/ml, and 50 g of streptomycin sulfate/ml in a 37°C tissue culture incubator. ␣IIbC5 cells from an ␣IIb integrin subunit-expressing CHO cell line were created by transfecting CHO cells with CDM8 vectors encoding the ␣IIb integrin subunit along with CD-neomycin plasmid. Stable colonies were selected with neomycin for 2 weeks. The pooled stable colonies were subsequently infected with an adenovirus encoding ␤3 integrin subunit and sorted for single cells expressing ␣IIb␤3 integrin. Each individual stable clone was then examined for the absence of ␤3 subunit to make sure that the adenovirus had not integrated into the genome. The clones that do not express ␤3 subunit were then transiently transfected with a plasmid that encodes ␤3 and tested for expression of ␣IIb␤3. The clones that revealed normal expression of ␣IIb␤3 after transient expression of ␤3 subunit were then used for further studies. The anti-␣IIb␤3 antibodies D57, PAC1, and anti-LIBS6 as well as Ro43-5054, an ␣IIb␤3-specifc peptide-mimetic competitive inhibitor, have been described previously (14 -17). The D57 antibody was biotinylated with biotin-N-hydroxy-succinimide (B-D57, Sigma) according to the manufacturer's instructions.
Construction of Random Mutagenic cDNA Libraries of ␤3 Integrin Subunit-To facilitate construction of random mutagenic cDNA libraries of transmembrane and cytoplasmic domains of ␤3 integrin subunit, a SphI restriction enzyme site was created via a silent mutation at nucleotide 1899 (C3 A) of ␤3 cDNA sequence using site-directed mutagenesis. The SphI site-containing full-length ␤3 cDNA was then subcloned into a hygromycin-resistant derivative of CDM8 expression vector (Invitrogen), CDHYG, to create a wide-type ␤3-expressing vector, CDHYG3A.SphI. Transiently transfection of this construct in a ␣IIbexpressing CHO cell line, ␣IIbC5, or co-transfection with wild-type ␣IIb cDNA in CHO cells resulted in a protein product with properties identical to those of wild-type ␣IIb␤3. Specifically, the mutant species was immunoprecipitated by D57 (data not shown), an antibody against the extracellular domains of ␣IIb␤3, and was found to be in the low affinity state because it failed to bind PAC1, a monoclonal antibody specific for the high affinity state of ␣IIb␤3. Furthermore, the SphI construct gained PAC1 binding ability with treatment of an ␣IIb␤3-activating antibody anti-LIBS6 but failed to bind PAC1 in the presence of Ro43-5054, an ␣IIb␤3-specifc peptide-mimetic competitive inhibitor (data not shown). Thus, this silent mutation does not interfere with the normal expression or function of ␤3 integrin subunit. The spiked megaprimer method was chosen to generate random mutagenic cDNA libraries of ␤3 cytoplasmic and transmembrane domains (18). Specifically, the ␤3 transmembrane and cytoplasmic domains were divided into four "windows" of 66 nucleotides, each with a nine-nucleotide overlap between windows ( Fig. 1). Spiked oligonucleotide primers corresponding to each window ( Fig. 1) were synthesized with a contamination level of 1.5% incorrect phosphoramidites, i.e. 0.5% of each of the three other bases. PCR was performed using the spiked primer ( Fig. 2A, S.M.P.) for each window and a 3Ј-reverse primer that contains an XbaI restriction enzyme site ( Fig. 2A, R.P.). For a second round of PCR reactions ( Fig. 2A, PCR-II), the PCR-I product (megaprimer) was used as a reverse primer. PCR reaction was performed with the megaprimer and a forward primer that contains the SphI site ( Fig. 2A, F.P.). Products from PCR-II were then subcloned into SphI-XbaI sites of CDHYG3A.SphI. The efficiency of random mutagenesis was assessed by cDNA sequencing of 10 randomly picked transformants from each window. cDNA sequencing indicated that ϳ70% inserts in each window contained 1, 2, or 3 point mutations, which is in the reasonable range of efficiency for random mutagenesis. Theoretically, for a window of 66 nucleotides, 150,000 transformants should cover 99% of the possible one-base changes and ϳ75% of the possible two-base changes. Therefore, a random mutagenic cDNA library containing ϳ200,000 transformants for each window of transmembrane and cytoplasmic domains of ␤3 subunit was constructed by large-scale preparation of transformants and used for subsequent transfection into ␣IIbC5 cells and identification of activating mutations in the ␤3 subunit.
Site-directed Mutagenesis-Site-directed mutations in both the ␣IIb and ␤3 subunits were generated using the QuikChange mutagenesis kit (Stratagene) and pCDM8 vectors containing the integrin subunits. Mutants were confirmed by DNA sequencing.
Flow Cytometry-Random mutagenic cDNA library corresponding to each window of ␤3 transmembrane and cytoplasmic domains was transfected into ␣IIC5 cells by electroporation. Stable colonies were selected for 2 weeks in the presence of both hygromycin (750 g/ml) and neomycin (750 g/ml). Approximately 5 ϫ 10 6 cells from pooled stable colonies from each window were then individually sorted on a FACStar Plus (BD Biosciences) using two-color flow cytometry. The biotinylated monoclonal antibody, B-D57, was used to detect the expression of ␣IIb␤3, whereas PAC1, an activation-specific monoclonal IgM antibody was used to assess the activation state of the ␣IIb␤3 integrin. For single cell sorting, the pooled stable colonies from each window (ϳ5 ϫ 10 6 cells/each window) were resuspended by treatment with trypsin and double-stained as described previously (19,20). Rare cells that exhibited both bright phycoerythrin staining (D57) and fluorescein isothiocyanate staining (PAC1) were individually sorted into 96-well plates.
FACS analysis of isolated clonal cell lines was performed on a FACScan using both B-D57 and PAC1 antibodies as described previously (19,20). PAC1 staining in the presence of Ro43-5054 (2 M) was used to estimate nonspecific binding. In some cases, treatment with anti-LIBS6 was used to estimate maximal PAC1 binding because anti-LIBS6 directly induces ␣IIb␤3 binding to PAC1 regardless of the FIG. 1. "Windows" for random mutagenic analysis of transmembrane and cytoplasmic domains of the ␤3 integrin subunit. The coding region for the ␤3 transmembrane and cytoplasmic domains was divided into four windows. Each window contains 66 nucleotides corresponding to the amino acid sequences with a nine-nucleotide overlap between windows.
For retransfection analysis, ␣IIbC5 cells were transfected with plasmid CDHYG3A.SphI, encoding wild-type or mutant ␤3, using Lipofectamine (Invitrogen) following the manufacturer's instructions. 48 h after transfection, the transfected cells were stained with B-D57 and PAC1 in the presence and absence of Ro43-5054 and FACS analysis was performed as described above.
Cytometric analysis of the site-directed mutations was done by cotransfection of the pCDM8 plasmids containing the ␣IIb and ␤3 subunits with Plus reagent and Lipofectamine. 48 h after transfection, the transfected cells were stained with B-D57 and PAC1 in the presence and absence of Ro43-5054 and anti-LIBS6 and FACS analysis was performed as described above. Activation index was calculated by using the formula ( binding in the presence of Ro43-5054, and F max ϭ PAC1 binding in the presence of anti-LIBS6. Reverse Transcriptase-PCR, Subcloning, and Sequencing of ␤3 Integrin Subunit-Total cellular RNA from each individual cell line was isolated using TRIzol (Invitrogen). cDNA was synthesized using oligo(dT) primer and the cDNA Cycle kit (Invitrogen), and PCR was performed following the manufacturer's instructions. PCR products were digested with restriction enzymes SphI and XbaI to create a fragment of ϳ550 bp. This SphI-XbaI fragment was then subcloned into the SphI-XbaI sites of CDHYG3A. SphI vector and sequenced using primers derived from CDHYG3A.SphI plasmid but outside of the SphI-XbaI region. To eliminate possible PCR errors, at least four clones were sequenced for each mutant cell line.
Computer Simulations-The modeling procedures for TM helix oligomerization were described elsewhere (21). The TMD sequences of integrin ␣IIb and ␤3 subunits were built into uniform ␣-helices having the backbone torsion angles of ϭ Ϫ65°and ϭ Ϫ40°. Their  I  288  91  7  II  468  192  22  III  65  5  0  IV  30  0  0 FIG. 2. Polymerase chain reactions for construction of random mutagenic cDNA libraries of the ␤3 transmembrane and cytoplasmic domain. A, schematic presentation of the partial ␤3 cDNA sequence. F.P. and R.P. represent forward primer and reverse primer, respectively, that were derived from the ␤3 sequence and used in the PCR reactions described in B and C. Asterisk represents the stop codon of ␤3 sequence. Newly created SphI site in the ␤3 sequence and XbaI site present in the vector sequence are indicated. B, schematic presentation of two consecutive PCR reactions for construction of random mutagenic cDNA library for each window of the ␤3 transmembrane and cytoplasmic domains. The position of each window is indicated. C, PCR product (Fig. 3B, PCR-I) for each window. PCR-I reaction was performed as described under "Experimental Procedures." 10 l of PCR product from each "window" was separated on 1% agarose gel and stained with ethidium bromide. The sizes of cDNAs are indicated. TMD sequences of ␣IIb and ␤3 were aligned based on the glycosylmapping data (22,23). Side chain rotamers were chosen using the backbonedependent rotamer library program SCWRL (24). Four hundred potential helix packings were first generated using a Monte Carlo search procedure as described previously (21). The ␣IIb and ␤3 TMD dimeric structures then were filtered to remove the structures incompatible with the bilayer constraints. We then clustered the remaining structures by C␣ root mean square distances using NMR CLUSTER (25), which resulted in two equally populated clusters: one with a crossing point near the N terminus and the other with a crossing point close to C terminus. Both models were evaluated for consistency with the experimental results (see below).

Random Mutagenesis Identifies Novel Integrin Activating Mutations in the TM and Membrane-proximal Cytoplasmic
Regions of ␤3-To generate the ␤3 random mutants, we used CDHYG3A.SphI as a template and spiked oligonucleotide primers corresponding to four overlapping windows that cover the entire ␤3 intracellular domain (Fig. 1). Using the protocol outlined (see "Experimental Procedure" and Fig. 2), a cDNA library containing ϳ200,000 transformants for each window was constructed by large-scale preparation of transformants and used for subsequent transfection into an ␣IIb-expressing cell line (␣IIbC5 cells) and identification of activating mutations in the ␤3 subunit. We developed a library of stable cell lines and selected activated ␣IIb␤3 integrin mutants by their binding to PAC1, an antibody specific for activated ␣IIb␤3 (Fig.  3A, R2) (15). Using this approach, we isolated 91 and 192 cell lines bearing activated ␣IIb␤3 from windows I and II, respectively (Table I). In contrast, only five cell lines were isolated from either window III or window IV (Table I). Thus, mutations in the membrane-distal segments of the ␤3 cytoplasmic domain encoded by regions III and IV were less likely to activate ␣IIb␤3 integrin.
The cell lines expressing activated integrin ␣IIb␤3 could have arisen as a consequence of mutations within the target window from adventitious mutations elsewhere in the integrin or from mutations in genes that indirectly control integrin activation. To identify mutations in the target window that activated ␣IIb␤3, we sequenced cDNA clones obtained from ␤3 reverse transcriptase-PCR amplicons spanning this region. To confirm that sequenced mutations were responsible for integrin activation, these amplicons were used to replace this region in wild-type ␤3. The resulting plasmids were transfected into ␣IIbC5 cells and tested for PAC1 binding in the presence and absence of Ro43-5054. When the transiently expressed mutant integrin was able to bind PAC1 and this PAC1 binding was inhibited by Ro43-5054, we concluded that the mutation was responsible for the activation of ␣IIb␤3. A total of 25 unique mutations in the transmembrane and membrane-proximal region of cytoplasmic domain of ␤3 subunit thus were identified, and there were multiple examples of the same mutations present in different clonal cell lines (Fig. 4). 13 of them were found N-terminal of Lys 716 in the presumptive transmembrane domain, and 12 were in the membrane-proximal region of cytoplasmic domain (Fig. 5). No mutation was identified that only affected the region of the cytoplasmic domain C-terminal of Asp 723 (Fig. 5). The absence of activating mutations C-terminal of Asp 723 indicates that the C terminus of ␤3 is not involved in the maintenance of the low affinity state of the integrin. The presence of such mutations in the membraneproximal and transmembrane domains suggests that these sites are important in regulating integrin activation.
Activating Mutations Are Predicted to Alter the TM Helix Length-We had previously established the importance of the membrane-proximal domains of the ␣ and ␤ subunits in regulating integrin activation (20). Consequently, we focused our attention on the numerous mutations in the transmembrane domain. Many of these mutations would be predicted to shorten the ␤-subunit TM helix (i.e. fewer residues embedded in the membrane) by deletion of one or more residues or by introduction of a charged residue (Table II). Indeed, the majority of the mutations (9 of 13) would be expected to disrupt or shorten the TM helix. Previous glycosylation-mapping studies (22,23) indicated that activating mutations in the membrane-proximal domain can shorten the TM helix. The present results extend those findings by showing that such mutations throughout the TM helix activate the integrin.

Mutagenesis of Predicted TM Packing Residues Activates
␣IIb␤3-A subset (4 of 13) of the membrane-embedded activating mutations had no obvious effect on TM helical length. This finding suggests that the TM segments help stabilize the inactive state through sequence-specific interactions. To investigate this possibility, we used a Monte Carlo simulation method (21) to produce a first approximation of the intersubunit packing of integrin TM domains. We caution that this method assumes idealized rigid ␣ helices and disregards potential changes in membrane insertion. However, several reports indicate that this protocol does yield models that conform well to known structural and functional data (21,(27)(28)(29). Here, the resulting output predicted two alternative structures, one with the TM helices packing near the C termini (Fig. 6, Structure A) and another with the helices packing close to the N termini (Structure B).
The random mutagenesis data focused our attention on Structure A, because all of the uncharged activating point mutations clustered in the C-terminal region of the ␤3 TM domain (G708S, A711T,T720A, and I714T) (Fig. 5). Indeed, each of these mutations affected a residue predicted to be a helical contact in Structure A (Fig. 6B). Three additional sitedirected mutants were employed to test the hypothesis that the interhelical interface predicted in Structure A functioned as a clasp to maintain the integrin in the low affinity state (vida infra).
The ␤3(G708S) mutation was strongly activating. Previous studies (30) have shown that the introduction of an Asn at this position activated ␣IIb␤3 and suggested that it did so by forming hydrogen bonds that favor ␤3 homo-oligomerization. The Ser substitution could also, in principle, lead to enhanced ␤3-␤3 interactions through such a mechanism. However, the weakly polar nature of the Ser side chain coupled with the observation that insertion of the bulky aliphatic Ile residue at this position (␤3(G708I) mutation) was strongly activating suggests that a Gly residue is strictly required at this position for efficient ␣/␤ TM packing (Fig. 7, A-C).
The packing motif in Structure A allowed us to identify an additional Van der Waals packing residue that helps stabilize the inactive state, because the apolar to apolar I704A mutation resulted in an activated integrin (Fig. 7, B and C). Also, the Structure A model predicted that ␣IIb TM residues would pack against the identified ␤3 residues. Accordingly, we substituted a bulky Ile residue for ␣IIb Thr 981 , the residue predicted to pack against ␤3 Gly 708 (Fig. 6B). The observation that this mutant activated ␣IIb␤3 supports the packing of Thr 981 against Gly 708 , an interaction that would be stabilized by both Van der Waals packing and a potential C␣/hydroxyl hydrogen bond. Overall, both random mutagenesis and site-directed mutagenesis (Fig. 7C) support the hypothesis that the specific packing of the C-terminal portions of ␣IIb and ␤3 transmem-brane helices against each other maintains the low affinity state of integrin ␣IIb␤3. Activation by mutagenically dissociating the integrin TM FIG. 6. Dimer models of integrin ␣II and ␤3 subunit TMD. A, Structure A, calculated model shows a left-handed helix-crossing angle of 30°with crossing point near C terminus. Packing residues are colored on the dimer model and indicated on the TMD sequence. From the packing residues, Gly 708 mediates the most-close packing. Alternative packing is shown in Structure B in which Model structure shows a left-handed helix-crossing angle of 40°with crossing point near N terminus. B, interhelical residues involved in Structure A. ␣IIb residues are highlighted in green, and ␤3 residues are highlighted in red.

FIG. 7.
Site-direct mutagenesis supports simulation Structure A as the inactive state conformation. A, CHO cells were transfected with plasmid coding for wild-type (WT) and mutant versions of ␣IIb and ␤3. Shown here a representative dot plots of FACS analysis for cells transfected with wild-type and G708I mutant integrin. Harvested cells were stained for integrin expression (D57) and activated ␣IIb␤3 (PAC1). B, activation indices for wild-type and site-directed mutants ␤3(I704A), ␤3(G708I), and ␣IIb(T981I) were calculated by measuring PAC1 staining in the presence or absence of anti-LIBS6 and Ro43-5054 (see "Experimental Procedures"). C, random and site-directed activating mutants map to interhelical residues outlined in computer simulation Structure A (underlined amino acids). Sites of activating random and site-directed mutations are shown with rectangles and circles, respectively. helices is consistent with reports that suggest that activation is associated with separation of the cytoplasmic domains (2,6). However, the unbiased random mutagenesis approach identified a preponderance of activating TM domain mutations predicted to shorten the ␤3 TM domain, indicating that TM helix shortening can also lead to integrin activation. In agreement with this mechanism, previous glycosylation mapping studies suggested that the membrane-proximal domains of the ␣ and ␤ subunits can reside with the membrane bilayers and that certain activating mutations in this region (22,23) result in a shortened TM domain.
How might shortening the TM domain lead to disruption of intracellular ␣/␤ interactions and consequent integrin activation? To avoid hydrophobic mismatch with the fixed width of the membrane bilayer, a shortened TM helix would change its membrane tilt angle and associated register with neighboring helices (31). Because helix-helix packing is dependent on specific crossing angles and specific in-register side chain arrays, changes to TM helical length would break the proposed clasp. Previous observations showing that ␣IIb sequences with a shortened TM segment (7) lost the ability to induce a periodic disulfide cross-linking pattern of the ␣IIb and ␤3 transmembrane helices support this notion. In addition, the inactive state intersubunit interactions at membrane-proximal level could cooperate with the TM packing to help maintain the ␣/␤ association. Importantly, talin has been shown to bind to the membrane-proximal region (8,32,33), an event that appears to be important for ␣IIb␤3 activation (33). This interaction could contribute to the physiological activation of integrins in two separate but related ways. First, one consequence of talin binding to this region would be to displace the membrane-proximal domain from the bilayer, thereby shortening the TM domain. As noted above, this process would probably lead to separation of the intracellular domains. Talin binding could also directly disrupt the cooperative membrane proximal/TM clasp by breaking the Arg 995 -Asp 723 salt bridge (8) and associated intersubunit interactions.
Our findings also have implications for integrin-mediated biochemical signals that control cell shape, cell migration, proliferation, and survival. The capacity of integrins to deliver such signals depends on their occupancy with resultant conformational change in the integrin (34) in combination with receptor clustering (35,36). These conformational changes are associated with a dramatic change in the quaternary structure of the integrin, resulting in a switch from a "bent" conformation observed in the crystal structure (37) to an extended one (38) that features a C-terminal separation (39) that would disrupt the TM helical packing proposed here. This disruption could lead to the changes in the intracellular interactions of occupied integrins manifested by focal adhesion targeting and transdominant inhibition (34,40). Furthermore, the work of Li et al. (10) shows that isolated integrin ␣ and ␤ TM peptides homooligomerize, a process that could contribute to integrin clustering. This finding suggests a sequential model in which ␣/␤ transmembrane separation might then be followed by homooligomerization to favor receptor clustering (10,26). Strikingly, the mutational studies of Li et al. (30) identify ␤3 Gly 708 as an important packing residue for ␤3 homo-oligomerization and suggest that homo-oligomerization may occur after TM separation. Our findings indicating that Gly 708 also participates in an ␣/␤ interaction that regulates activation lends additional credence to this hypothesis and suggests that the TM helix packing interface proposed here is a nexus for bidirectional transmembrane signaling through integrins.
Note Added in Proof-Recently, two other groups (41,42) have reported studies supporting the hypothesis that heterodimeric transmembrane domain-packing interactions stabilize the integrin-inactive state.