Interactions of platelet integrin α_ΙΙb and β₃ transmembrane domains in mammalian cell membranes and their role in integrin activation

Integrin-mediated cell adhesion modulates signaling pathways that control many biologic functions,¹  and these signals involve both integrin clustering and integrin occupancy² ; the latter changes integrin conformation, leading to allosteric rearrangements that propagate across the membrane and modify intracellular interactions.³  A second form of integrin signaling, initiated intracellularly, leads to increased affinity for ligands, a process termed integrin activation.¹  Both forms of integrin signaling are central to the functions of platelet integrin α_IIbβ₃ (GPIIb-IIIa) in hemostasis and thrombosis.⁴  Integrin activation is also essential for functions such as inflammation and assembly of the extracellular matrix. Integrins are noncovalent α-β heterodimeric, type-1 transmembrane (TM) receptors formed from combinations of 18 α and 8 β subunits. Many studies indirectly implicate interactions between integrin α and β TM domains (TMDs) and cytoplasmic domains (tails) in both forms of integrin signaling^5-11 ; however, studies to directly identify such interactions have met with contradictory results.

Mutational studies of platelet integrin α_IIbβ₃ (GPIIb-IIIa), based on sequence alignments, suggested an interaction between β₃ Asp⁷²³ and α_IIb Arg⁹⁹⁵.⁵  Charge reversal mutation of either residue resulted in constitutive bidirectional integrin signaling (ie, integrin activation and constitutive phosphorylation of pp125^FAK); however, a double-charge reversal in α_IIb(R995D)β₃(D723R) did not exhibit constitutive signaling, suggesting that a salt bridge between these residues was a defined structural constraint that limited the integrin signaling.⁵  A spontaneous activating mutation (β₃(D723H)) found in patients also suggests the importance of the β₃ Asp⁷²³ residue for stabilization of the inactive state of the integrin.¹²  This constraint gained support from elegant protein engineering studies in which clasping the cytoplasmic domains or TMDs together inhibited integrin activation,⁹  and joining of the α and β TMDs with disulfide bonds limited bidirectional integrin signaling.⁷  Despite this mutational data, our initial efforts to identify interactions of isolated α and β integrin tails by nuclear magnetic resonance (NMR) spectroscopy were unsuccessful in aqueous solution.¹³  Two laboratories reported NMR studies of these tails that suggested the existence of such interactions, albeit different structures for the αβ dimer were reported.^14,15  Structures of the individual α_IIb and β₃ TMDs in phospholipid bicelles^16,17  show that the β₃ TMD adopts an elongated, tilted membrane helix¹⁷  and the α_IIb TMD folds into a short, straight helix, followed by a surprising backbone reversal that packs Phe⁹⁹² and Phe⁹⁹³ against the TM helix.¹⁶  This structure demonstrated an unexpected complexity in the α_IIb TMD and membrane-proximal cytoplasmic domain that was not observed in the studies of cytoplasmic domains in aqueous solution,^14,15  thus calling NMR studies that suggested the interactions into question.

Whereas mutational studies and molecular modeling have strongly suggested that integrin αβ TMD interactions are important in regulating integrin signaling,^{8,10,11,18,19}  efforts to identify direct interactions between α and β TMDs (either biochemically or genetically) have had contradictory results.^20-23  Indeed, αα and ββ interactions were proposed to make a major contribution to the clustering that occurs in integrin signaling²² ; however, later studies have strongly challenged this idea.^8,11  Furthermore, a 20 Å reconstruction of detergent-solubilized integrin α_IIbβ₃ revealed a cylindrical density interpreted as a TM segment that suggests that the TMDs are associated in a parallel, α-helical coiled-coil²⁴ ; however, as noted above, higher resolution structures of the individual α_IIbβ₃ TMDs in a lipid environment are not consistent with a purely coiled-coil architecture. In this study, we use a novel mini-integrin affinity capture approach to establish the preferential heteromeric interaction of integrin α_IIb and β₃ TMDs and cytoplasmic domains in mammalian cells, and show by NMR spectroscopy that heterodimerization takes place analogously in small bicelle model membranes. Thus, we show that integrin TMDs and cytoplasmic domains preferentially interact in a heterodimeric rather than homodimeric fashion in mammalian cell membranes, interaction between the highly conserved β₃ Asp⁷²³ and α_IIb Arg⁹⁹⁵ stabilizes this interaction, and talin binding to the β₃ cytoplasmic domain disrupts it.

Tac-α_IIb^TM, Tac-β₃^TM were generated by ligation of polymerase chain reaction (PCR)–generated fusion sequences consisting of extracellular domain of Tac and TM tail regions of each integrin subunit (Figure 1A) into pcDNA3.1 (Invitrogen, Carlsbad, CA). For the construction of α_IIb^TM- TAP (tandem affinity purification), the sequences of the preprotrypsin leader sequence followed by 3 repeats of FLAG sequence (Sigma-Aldrich, St Louis, MO) were first cloned into pcDNA3.1, and then the PCR-generated fusion sequences containing α_IIb^TM and TAP were ligated. Point mutations were performed using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Anti–Tac 7G7B6 (ATCC, Manassas, VA), anti–Tac N19 (Santa Cruz Biotechnology, Santa Cruz, CA), and anti–FLAG M2 (Sigma-Aldrich) were obtained commercially. Rabbit polyclonal anti-β₃ antibody (Rb8053), mouse monoclonal antibody specific for the human integrin α_IIb subunit (PMI-1), activation-specific anti-α_IIbβ₃ antibody (PAC1), and α_IIbβ₃ activating antibody (anti-LIBS6) have been described previously.²⁶  Chinese hamster ovary (CHO) cells and CHO cells expressing recombinant α_IIbβ₃ (A5) were maintained, as described previously.²⁶  Lipofectamine and Lipofectamine Plus reagents (Invitrogen) were used according to the manufacturer's recommendation for transient transfections.

Twenty-four hours after transfection, cells were lysed with CHAPS (3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate) lysis buffer (20 mM HEPES [N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid] pH 7.4, 1% CHAPS, 150 mM NaCl, 2 mM CaCl₂, and EDTA [ethylenediaminetetraacetic acid]–free protease inhibitor mixture [Roche, Basel, Switzerland]) and clarified by centrifugation at 14 000 rpm for 15 minutes, and then the clarified lysates were incubated with calmodulin Sepharose (GE Healthcare, Piscataway, NJ) for 2 hours at 4°C. Bound proteins were eluted with sodium dodecyl sulfate (SDS) reducing sample buffer, subjected to SDS–polyacrylamide gel electrophoresis (PAGE), and analyzed by Western blot. For capturing biotinylated intact integrin subunits, A5 cells expressing TAP constructs were detached and surface proteins were biotinylated using EZ-Link Sulfo-NHS-Biotin (Thermo Scientific, Rockford, IL) according to the manufacturer's recommendation. The cells were washed twice with Tris-buffered saline (TBS; pH 8.4), and incubated for 30 minutes at 37°C with TBS containing 5 mM EDTA. Resulting cells were washed with HEPES-buffered saline (20 mM HEPES, 150 mM NaCl, pH 7.4) before being lysed with CHAPS lysis buffer. Lysates were incubated with calmodulin-beads to capture TAP constructs for 2 hours, bound proteins were eluted with 10 mM EDTA, and then the eluates were further incubated with NeutrAvidin agarose resin (Thermo Scientific) overnight to capture biotinylated proteins. Bound proteins were eluted with SDS reducing sample buffer, subjected to SDS-PAGE, and analyzed by Western blot.

Peptides encompassing human integrin α_IIb(Ala⁹⁵⁸-Pro⁹⁹⁸) and β₃(Pro⁶⁸⁵-Phe⁷²⁷) residues, respectively, including β₃(Cys⁶⁸⁷Ser), were prepared as described previously.^16,17  In addition, an α_IIb peptide incorporating Arg⁹⁹⁵Ala and a β₃ peptide with an Asp⁷²³Ala substitution were prepared analogously. The peptides were reconstituted in 385 mM 1,2-dihexanoyl-sn-glycero-3-phosphocholine, 83 mM 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-choline, 41 mM 1-palmitoyl-2-oleoyl-sn-glycero-3-[phospho-l-serine], 25 mM HEPES·NaOH, pH 7.4, 6% D₂O, and 0.02% wt/vol NaN₃. Transverse relaxation optimized–heteronuclear single quantum coherence (TROSY-HSQC) NMR experiments were conducted on a cryoprobe-equipped Bruker Avance 700 spectrometer at 23°C.

PAC1-binding assay and fibronectin-binding assay were performed essentially as described.^26,27  In brief, 1 day after transfection, suspended cells were incubated with 7G7B6 (anti-Tac) in combination with either PAC1 or biotinylated glutathione S-transferase-FN9-11, followed by staining with fluorescein isothiocyanate (FITC)–conjugated anti–mouse immunoglobulin (Ig) G and with R-phycoerythrin (PE)–conjugated anti–mouse IgM or PE-conjugated streptavidin. Five minutes before analysis, propidium iodide (PI) was added, and PI-negative live cells were analyzed on FACSCalibur (BD Biosciences, San Jose, CA). The data were analyzed using MATLAB R2007a software to calculate the geometric means and generate dot plots.

A5 cells expressing Tac-α_IIb^TM or Tac-α_IIb were detached and stained with anti–Tac N19 antibody and PAC1 in the presence or absence of anti-LIBS6, as indicated, followed by staining with FITC-conjugated anti–rabbit IgG and rhodamine-conjugated anti–mouse IgM. The stained cells were fixed, mounted in Prolong Gold (Invitrogen), and observed under a Nikon Eclipse TE2000-U inverted microscope. Images were acquired using 60×/1.4 NA oil-objective lens with 1.5× intermediate magnification, a Coolsnap HQ camera (Photometrics, Tucson, AZ), and QED InVivo imaging software (Media Cybernetics, Silver Spring, MD) at room temperature. Postacquisition image analysis was performed using ImageJ (National Institutes of Health, Bethesda, MD).

To test whether the TMDs of α_IIb and β₃ interact with each other, we constructed an α_IIb mini-integrin containing the TMD and cytoplasmic tail of α_IIb (Figure 1A) joined to an N-terminal flag tag for detection and a C-terminal TAP tag²⁸  for rapid and efficient purification (Figure 1B). We expressed this α_IIb^TM-TAP bait in combination with preys comprising the extracellular domain of the Tac (interleukin-2 receptor α) joined to the TMD and tail of α_IIb (Tac-α_IIb^TM) or β₃ (Tac-β₃^TM; Figure 1B). The cells were lysed, and baits were captured using calmodulin beads. In the reaction, approximately 20% of expressed baits were usually captured. We detected bound prey by Western blotting with anti-Tac antibody and found that α_IIb^TM bait bound preferentially to Tac-β₃^TM rather than Tac-α_IIb^TM (17-fold more in quantification; Figure 1C). This interaction required the β₃ TMD, because a prey containing the Tac extracellular and TMD joined to the β₃ tail failed to bind to the α_IIb^TM bait (Figure S1A, available on the Blood website; see the Supplemental Materials link at the top of the online article). We noted multiple bands for each Tac construct and enzymatic deglycosylation and cell-surface biotinylation experiments showed that the top band observed in Figure 1C is a glycosylated form found predominantly on the cell surface, whereas the bottom band is a nonglycosylated intracellular form (Figure S1B,C). When we surface labeled the cells, we again found that α_IIb^TM-TAP bait interacts with Tac-β₃^TM (Figure 1D left panels) and only the top band was observed. Conversely, bait containing the TMD and tail of β₃ preferentially bound the Tac-α_IIb^TM rather than Tac-β₃^TM on the cell surface, although subtle ββ homomeric interactions were also observed (Figure 1D right panels). Thus, we observed preferential heteromeric interactions between α_IIb and β₃ TMD-tail in both cell surface and intracellular membranes.

The foregoing results showed preferential heteromeric interactions between baits and preys containing the TMDs and cytoplasmic domains of integrin α_IIbβ₃, and that the interaction required the β₃ TMD. To test the role of the α_IIb TMD, we changed 2 Gly in the α_IIb TMD GXXXG motif to Leu (Figure 1A). This motif is buried in the hydrophobic core of the lipid bilayer and is predicted to be important in αβ TMD packing.^8,19  As a second test of specificity, we deleted the membrane-proximal GFFKR motif; both this deletion and the Leu substitutions activate integrin α_IIbβ₃.⁸  Both of these mutations markedly reduced interaction of α_IIb bait with β₃ prey (Figure 1E). Thus, the bait-prey interactions detected in this study are mediated by specific packing interactions of the TMDs. The strong preference for heteromeric αβ interactions in mammalian cell membranes is in sharp contrast to the homodimeric and trimeric interactions previously observed in detergent micelles,²¹  in electrophoresis,^21,22  and in the Escherichia coli inner membrane.²³  Schneider and Engelman²⁰  did observe promiscuous weak heterodimeric interactions of central hydrophobic 17-residue fragments of the integrin TMDs using transcriptional repression of LacZ by dimer formation in E coli membranes; those authors observed extensive homo-oligomerization as well. Thus, the present data provide the first direct evidence for the heteromeric association of α_IIbβ₃ TMD-tails in mammalian cell membranes.

The plasma membrane is a complex environment; that is, it contains many proteins and lipids. Whereas this complexity is representative of the environment of integrin α_IIbβ₃ in blood platelets,²⁹  it raises the possibility of indirect α_IIb-β₃ interactions. Using NMR spectroscopy, we therefore assessed the association of peptides corresponding to α_IIb(Ala⁹⁵⁸-Pro⁹⁹⁸) and β₃(Pro⁶⁸⁵-Phe⁷²⁷), which contain the TMDs^16,17  and possible cytosolic heterodimerization sites in the controlled lipid environment of small bicelle model membranes. Within this environment, the presence of the α_IIb peptide led to the appearance of a second set of NMRs for the β₃ TMD residues in addition to the previously observed monomeric signals,^16,17  as exemplified for β₃(G702) and β₃(G708) (Figure 2A-C). This second set of resonances arises from α_IIb-β₃ association (heterodimerization), which creates new chemical environments for all α_IIbβ₃ TMD resonances (Figure S2A) as a consequence of direct α_IIb-β₃ contacts and indirect, propagated effects. For example, β₃(G708) is expected to form the dimerization interface, whereas β₃(G702) is peripheral and most likely experiences predominantly indirect, next-neighbors changes. To test the existence of the proposed α_IIb(R995)-β₃(D723) salt bridge and verify the specificity of the α_IIb-β₃ interaction, wild-type α_IIb TMD peptide was substituted by α_IIb(R995A).⁵  This mutant α_IIb peptide, whose backbone fold is indistinguishable from wild-type α_IIb (Figure S2B), failed to induce a significant second set of β₃ signals (Figure 2D), demonstrating that heterodimerization weakened considerably. Analogously, mutant β₃(D723A) peptide did not induce a second set of α_IIb resonances in contrast to wild-type peptide (data not shown). Thus, in the defined model membrane environment of phospholipid bicelles, specific TM α_IIb-β₃ interactions, leading to heterodimerization, were observed and were dependent on α_IIb(R995) and β₃(D723) electrostatic interactions.

To assess whether α_IIb(R995)-β₃(D723) electrostatic interaction also takes place in cell membranes, we introduced charge reversal mutations in these residues creating α_IIb(R995D) bait and β₃(D723R) prey. The combination of α_IIb(R995D) bait with β₃ prey, or of β₃(D723R) prey with α_IIb bait, resulted in marked reduction in the interaction (Figure 2E). The αβ interaction was largely rescued when a combination of α_IIb(R995D) bait and β₃(D723R) prey was used (Figure 2E). Thus, both NMR in synthetic membranes and affinity capture in mammalian cell membranes confirm the importance of the α_IIb(R995)-β₃(D723) interaction in stabilizing the association of the α_IIbβ₃ TMD-tails. Our previous inability to detect specific α_IIb(Arg⁹⁹⁵)-β₃(Asp⁷²³) contacts in aqueous solution¹³  indicates their weakness in the absence of a lipid milieu that provides a lowered dielectric constant and reduced solvent water concentration.

Talin binding to integrin β cytoplasmic domains is a final step in integrin activation.^30,31  Transfection of cells with the talin head domain (THD) inhibits Förster resonance energy transfer (FRET) between donors and acceptors fused to the cytoplasmic tails of α_L and β₂ integrins,⁶  indicating a change in the orientation and/or distance between the tails. Having developed a direct measure of the association of the integrin α and β TMD-tails, we asked whether talin binding can inhibit the interaction. THD reduced the α_IIb^TM-TAP-Tac-β₃^TM interaction by 40% relative to a β₃ binding-defective mutant THD, THD(W395A)³²  (Figure 3A). Conversely, THD did not block binding of Tac-β₃^TM(Y747A), a mutation that inhibits talin binding,³²  to α_IIb^TM-TAP (Figure 3B). In the absence of THD, Tac-β₃^TM(Y747A) also showed increased association with α_IIb^TM-TAP, probably due to the presence of endogenous talin in the CHO cells (data not shown). Thus, talin binding to the β₃ tail inhibits the interaction of the α_IIb and β₃ TMD-tails. The partial dissociation of the αβ complex by THD may be due to variation in the ratio of THD to Tac-β₃^TM and α_IIb^TM-TAP on a per cell basis in these triple transfection experiments.

It is notable that mutations in the α_IIb GXXXG motif, located near the outer leaflet of the membrane, and mutations that disrupt the salt bridge or talin binding, which would act near the inner leaflet of the membrane, both decrease the TMD-tail association. These results raise the intriguing possibility that both outer membrane interaction (including the α_IIb GXXXG motif) and inner membrane interaction (involving α_IIb(F992,933 and R995) and β₃(D723) are required for stabilization of the TMD interface and the inactive state of the integrin.

We reasoned that should these TMD-tails interact with native α_IIbβ₃, then they might disrupt the interaction of the endogenous α_IIb and β₃, leading to integrin activation. Indeed, overexpression of Tac-α_IIb^TM induced a dose-dependent increase in the binding of an activation-specific anti-α_IIbβ₃, PAC1,³³  whereas a construct lacking the α_IIb TMD (Tac-α_IIb) exhibited no such effect (Figure 4A). The Tac-α_IIb^TM failed to activate α₅β₁ (Figure S3), supporting the integrin class specificity of interactions of integrin TMDs.³⁴  Additional evidence for specificity of this effect was provided by mutants that disrupt the TMD-tail interaction of integrin α_IIb and β₃; Tac-α_IIb^TM constructs in which we deleted the GFFKR motif or mutated the 2 Gly in the GXXXG motif failed to activate α_IIbβ₃ (Figure 4B).

Expression of the free β₃ cytoplasmic domain inhibited integrin activation (Figure 4C) probably due to sequestration of endogenously expressed talin.³³  Addition of the β₃ TMD to the cytoplasmic domain led to reversal of the inhibitory effect and integrin activation at higher levels of expression (Figure 4C), possibly resulted from combination of inhibitory effect by sequestering talin and activating effect by binding to the TMDs of the intact integrin. Furthermore, the β₃(Y747A) mutation, which blocks talin binding,³²  increased the ability of the β₃ TMD-tail to activate integrin α_IIbβ₃ and reduced the inhibitory effect (Figure 4C). Thus, overexpression of the TMD tail of either α_IIb or β₃ activated the native integrin, and activation depended on the presence of the TMDs.

To examine the interaction of α_IIb TMD-tail with native α_IIbβ₃, we captured the α_IIb^TM-TAP bait, resulting in isolation of native α_IIbβ₃ (Figure 5A), whereas deletion of the GFFKR motif inhibited the association (Figure 5A). We also examined the association of Tac-α_IIb^TM with activated α_IIbβ₃ in cells. Most PAC1-positive clusters of activated α_IIbβ₃ also contained the Tac-α_IIb^TM (Figure 5B). Cells transfected with Tac-α_IIb, which lacks the TMDs, failed to stain with PAC1; however, when the α_IIbβ₃ was activated with an activating antibody, anti-LIBS6,³⁵  Tac-α_IIb was not associated with the clusters of activated integrin (Figure 5B). Thus, the α_IIb TMD tail interacts with native α_IIbβ₃, and this physical association activates the integrin.

Consistent with the findings described above, a peptide containing a partial sequence of α_IIb TMD (Trp⁹⁶⁸-Lys⁹⁸⁹) can bind and activate α_IIbβ₃³⁶ ; however, this effect was ascribed to a homodimeric interaction of the α_IIb peptide with the full-length α_IIb. To assess which subunit of full-length α_IIbβ₃ binds to the α_IIb TMD-tail, we used EDTA to dissociate the subunits and verified the dissociation by the loss of binding of a complex-specific antibody (Figure 5C). Because the EDTA treatment can affect only cell surface–expressed integrin, we surface biotinylated cells expressing α_IIbβ₃ in combination with α_IIb^TM-TAP to exclude the intracellular integrin from this assay, treated the cells with EDTA to dissociate the α_IIbβ₃ on cell surface, and then captured α_IIb^TM-TAP with immobilized calmodulin (Figure S4). Biotinylated surface proteins were then affinity purified with NeutrAvidin beads and detected by antibody against α_IIb (PMI-1) and β₃ (Rb8053). We found that β₃ subunit was captured by α_IIb^TM-TAP (Figure 5D top, lane 2), and the LXXXL and ΔGFFKR mutations markedly reduced the association (Figure 5D top, lanes 3,4). In sharp contrast, only nonspecific (as judged by equal interactions of wt, LXXXL, or ΔGFFKR baits and by the ratio of captured [Figure 5D lanes 2-4] to input of the cell surface–expressed integrin subunits [Figure 5D lanes 6-8] were observed with the α_IIb subunit (Figure 5D middle panel). This result shows that the α_IIb^TM-TAP preferentially captures full-length integrin β₃ in the presence of an equimolar amount of native full-length α_IIb. Thus, the unpaired α_IIb TMD-tail induces equilibrium shift in favor of activation of integrin α_IIbβ₃ by interacting with the β₃ subunit and separating the TMD-tail interaction of the intact integrin (Figure 5E).

Bidirectional TM integrin signaling is central to integrin-mediated biological functions. Mutational studies suggested that this signaling involves heterodimeric interactions of the TMD-tails; frustratingly, efforts to directly demonstrate these interactions produced conflicting results. In this study, we took advantage of the high efficiency and rapidity of TAP tag purification to show preferential heterodimeric interactions of the α_IIb and β₃ TMD-tails in mammalian cell membranes. Previous studies in E coli inner cell membranes showed primarily homodimeric interactions of the α_IIb TMDs; the variance from our results is most likely ascribable to the fact that we used a full-length cytoplasmic domain and included the GFFKR motif, shown in this study to be critical for stabilization of the association of the TMD-tail mini-integrins. Similarly, previous studies in detergent micelles suggested primarily homomeric α_IIb and β₃ TMD-tail interactions²¹ ; however, phospholipid bicelles are a more accurate representation of the bulk mammalian plasma membrane because they contain lipids in a bilayer arrangement. Indeed, embedding in dodecylphosphocholine micelles distorts the helical structure of the β₃ membrane-proximal domain,¹⁷  a region important in regulating integrin signaling. The remarkable agreement of biochemical and biophysical methods in identifying heterodimeric associations, combined with previous mutational data, provides compelling evidence for preferential heterodimeric TMD-tail interactions of α_IIbβ₃ in the lipid environment and the mammalian cell membrane. Furthermore, we now provide direct proof that talin binding disrupts the TMD-tail association and that an α_IIb(R995)β₃(D723) interaction stabilizes it. Moreover, the new approaches described in this study now enable analysis of TMD-tail interactions among many type I membrane proteins, such as tyrosine kinase growth factor receptors, immunoreceptors, or cytokine receptors that transduce signals as homo- and heteromultimers.

This work was supported by grants from the National Institutes of Health (Bethesda, MD) to M.H.G. (HL70784 and AR27214) and T.S.U. (HL089726). C.K. and T.-L.L. are recipients of postdoctoral fellowships from the American Heart Association (Dallas, TX).

National Institutes of Health

Contribution: C.K. designed and performed the mammalian cell experiments and wrote the paper; T.-L.L. designed and performed the NMR experiments; T.S.U. designed the NMR experiments and edited the paper; and M.H.G. designed the mammalian cell experiments and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Dr Mark H. Ginsberg, Department of Medicine, University of California San Diego, 9500 Gilman Dr, La Jolla, CA 92093; e-mail: mhginsberg@ucsd.edu.