Activation of Integrin αIIbß3 by Modulation of Transmembrane Helix Associations

See allHide authors and affiliations

Science  02 May 2003:
Vol. 300, Issue 5620, pp. 795-798
DOI: 10.1126/science.1079441


Transmembrane helices of integrin α and β subunits have been implicated in the regulation of integrin activity. Two mutations, glycine-708 to asparagine-708 (G708N)and methionine-701 to asparagine-701, in the transmembrane helix of the β3 subunit enabled integrin αIIbβ3 to constitutively bind soluble fibrinogen. Further characterization of the G708N mutant revealed that it induced αIIbβ3 clustering and constitutive phosphorylation of focal adhesion kinase. This mutation also enhanced the tendency of the transmembrane helix to form homotrimers. These results suggest that homomeric associations involving transmembrane domains provide a driving force for integrin activation. They also suggest a structural basis for the coincidence of integrin activation and clustering.

The mechanism regulating the activation state of integrins is emerging from structural investigations (1, 2). Electron microscopy of several integrins and crystallography of αvβ3 (35) indicate that integrins are composed of a globular ligand-binding head and two flexible rodlike stalks containing the carboxyl-terminal portions of the integrin α and β chains, respectively. In the αvβ3 crystal structure (5), the stalks are closely aligned but bend sharply half-way through their length, causing the globular head to project toward the cell surface. On the basis of electron microscopy and biochemical studies, Springer and co-workers have proposed that the bent conformation corresponds to an inactive conformation, which is converted to a more upright configuration upon activation (6). Though some details of this mechanism remain somewhat controversial (7, 8) and alternative mechanisms have been proposed, one feature appears common to a number of proposals: coincident with activation, the stalks splay apart (2, 3, 9, 10). Thus, the transmembrane-cytoplasmic domains of integrin α and β subunits, present at the ends of the stalks, are likely to be closely aligned in the inactive state and separated in the activated state.

Isolated α and β transmembrane helices from αIIbβ3 form homomeric dimers and trimers, respectively (11). These oligomers have been observed in both acidic and zwitterionic micelles and have affinities consistent with well-defined transmembrane homooligomeric bundles. On the basis of the crystal structure of αvβ3, homomeric interactions would only be sterically possible in the activated state in which the α and β stalks are splayed apart and removed from the central globular domain. This would thermodynamically or kinetically link integrin activation (6) and homo-oligomerization.

Thus, homomeric association of integrin transmembrane helices is poised to regulate integrin activity, and mutations that subtly affect homomeric affinity should increase the constitutive level of integrin activation. Addition of polar side chains, such as those of Asn, Asp, Glu, and Gln, to model transmembrane helices promotes their tendency to associate in biological membranes in vivo (12) and favor oligomerization by a relatively modest but energetically important 1 to 2 kcal/mol of helix (12, 13). We placed Asn at consecutive positions in 10 different β3 transmembrane domain mutants (Fig. 1A), with the aim of increasing the tendency of the β3 transmembrane helix to form trimers, thereby altering the αIIbβ3 activation state. Each mutant was co-expressed with wild-type αIIb in Chinese hamster ovary (CHO) cells that readily express recombinant αIIbβ3 in an inactive conformation (14). Expression of 7 of the 10 β3 mutants on the CHO cell surface was comparable to that of wild-type β3 (15). However, the Ala703 3 Asn703 (A703N), L706N, and I707N (16) mutants were expressed at substantially lower levels (fig. S1), suggesting that Asn at these positions in the β3 helix has a deleterious effect on either β3 and/or αIIbβ3 biosynthesis.

Fig. 1.

(A) Sequence of the β3 transmembrane helix. The amino acids that were replaced by Asn are underlined. (B) Comparison of constitutive and DTT-induced fibrinogen binding to CHO cells expressing wild-type (WT) human αIIbβ3 or G708N β3 mutant. FITC-fluorescence intensity, corresponding to fibrinogen binding, is shown on the y axis, and phycoerythrin (PE)-fluorescence intensity, corresponding to β3-specific mAb SSA6 binding and the level of β3 expression, is shown on the x axis. (C) Inhibition of constitutive FITC-fibrinogen binding to CHO cells expressing β3 G708N by 5 mM EDTA, 0.5 mM RGDS, and 50 μg/ml A2A9. (D) Constitutive fibrinogen binding to CHO cells expressing WT αIIbβ3 and the G702N and G708N β3 mutants as a function of fibrinogen concentration. (E) Constitutive fibrinogen binding to CHO cells expressing WT β3 and the indicated β3 mutants. The data are expressed as the ratio of αIIbβ3-expressing cells constitutively binding fibrinogen to αIIbβ3-expressing cells not binding fibrinogen determined from dot plots of two-color flow cytometry and were normalized to data obtained from the cells expressing wild-type αIIbβ3. The data shown are representative of three to six independent experiments.

We compared the ability of wild-type αIIbβ3 and the seven highly expressed αIIbβ3 mutants to bind soluble fibrinogen, both constitutively and after incubation with the known activator dithiothreitol (DTT) (14). The mutations did not disrupt αIIbβ3 function so that wild-type αIIbβ3 and each of the mutants readily bound soluble fibrinogen in the presence of DTT. However, there were significant differences in constitutive fibrinogen binding. The G708N mutant constitutively bound ≈fivefold more fibrinogen than wild-type αIIbβ3 (Fig. 1B), and analysis of the fluorescence-activated cell sorter (FACS) data revealed that approximately 30 to 50% of the G708N mutants were constitutively activated. Constitutive fibrinogen binding was prevented by the divalent chelator EDTA (17), the tetrapeptide RGDS (18), and the inhibitory monoclonal antibody (mAb) A2A9 (19), confirming that the fibrinogen was bound to αIIbβ3 (Fig. 1C). Moreover, fibrinogen binding to cells expressing G708N was saturable, whereas fibrinogen binding to cells expressing wild-type αIIbβ3 increased linearly as the concentration of fluorescein isothiocyanate (FITC)-fibrinogen increased (Fig. 1D). A double reciprocal plot of fibrinogen binding to G708N revealed a dissociation constant (Kd) of 180 nM, similar to the Kd's of 81 nM and 178 nM for fibrinogen binding to ADP-stimulated and epinephrine-stimulated platelets, respectively (20).

The M701N mutant, located one heptad apart from G708 and, hence, along the same face of the β3 transmembrane helix, constitutively bound ≈threefold more fibrinogen than wild-type αIIbβ3 (Fig. 1E). In contrast, the constitutive fibrinogen binding activity of G702N, a mutant whose side chain projects orthogonal to that of G708, as well as that of the other four β3 mutants, was similar to that of wild-type αIIbβ3 (Fig. 1E). Comparable results were observed when FITC-labeled PAC1, a mAb that exclusively recognizes the activated αIIbβ3 conformation (21), was substituted for FITC-fibrinogen (22). Thus, the introduction of an Asn side chain at an appropriate position on one face of the β3 transmembrane helix is sufficient to shift αIIbβ3 from an inactive to an active conformation.

Then, we used four experimental approaches to determine whether the αIIbβ3 activation induced by the G708N mutation is accompanied by αIIbβ3 oligomerization. First, cells expressing wild-type αIIbβ3 and the G708N and G702N mutants were fixed and stained sequentially with the β3-specific mAb SSA6 and FITC-labeled mouse antibody to immunoglobulin G (anti-mouse IgG). Fluorescence microscopy images of the stained cells were then compared with images in which wild-type αIIbβ3 was clustered by cross-linking αIIbβ3-bound SSA6 on unfixed cells with anti-mouse IgG. Wild-type αIIbβ3, clustered by cross-linking, was present in variably-sized fluorescent patches on the CHO cell surface, whereas wild-type αIIbβ3 and the G702N mutant formed a homogenous ring at the cell periphery (Fig. 2). No ring of fluorescence was present in cells expressing the G708N mutant; rather, there were patches of fluorescence similar to those on cells in which αIIbβ3 had been cross-linked. Thus, in the absence of ligands, the G708N mutation was sufficient by itself to induce the formation of αIIbβ3 clusters.

Fig. 2.

Detection of αIIbβ3 on the surface of CHO cells expressing WT αIIbβ3 and the G702N and G708N β3 by fluorescence microscopy. Cells were fixed with paraformaldehyde and were stained with the anti-β3 mAb SSA6 and FITC-labeled anti-mouse IgG (15). Cells in the row labeled WT+Ab were stained with SSA6 and FITC-labeled anti-mouse IgG before fixation to deliberately cluster αIIbβ3. Bar, 30 μm.

Second, we used a functional assay to monitor clustering. Integrin clustering in focal adhesions is accompanied by the tyrosine phosphorylation of focal adhesion kinase (FAK) (23, 24). As expected, FAK was extensively phosphorylated when CHO cells expressing wild-type αIIbβ3 and the G702N mutant were allowed to adhere to a fibrinogen-coated surface, but there was only minimal phosphorylation when these cells were placed in suspension (Fig. 3A). In contrast, FAK was extensively phosphorylated in CHO cells expressing the G708N mutant, regardless of whether the cells were adherent to fibrinogen or free in suspension. Thus, the G708N mutation promotes downstream signaling events known to depend on integrin clustering.

Fig. 3.

The effects of the G708N mutation. (A) Phosphorylation of FAK in CHO cells expressing WT αIIbβ3 and the G702N and G708N β3 mutants. The CHO cells were either adherent to immobilized fibrinogen or maintained in suspension. (Top) The extent of FAK phosphorylation detected by an anti-phosphotyrosine antibody. (Bottom) The amount of FAK detected by an anti-FAK antibody. (B) SDS-PAGE of peptides corresponding to the transmembrane helix of WT and G708N β3. The amount of peptide in 20 μl sample volume loaded per lane is indicated at the top of the gel. Two 29-residue transmembrane peptides with different oligomeric states were present as molecular weight markers. MS1 is a trimer, whereas MS1-N14A is a monomeric MS1 mutant (31). (C) Relative composition of oligomer species as a function of peptide/detergent molar fraction. The plots were calculated from analytical ultracentrifugation data for the wild-type and G708N β3 transmembrane helix in 8 mM C14-betaine at pH 7.4 (15). The observed species for each peptide are labeled in the plots.

Although these results are consistent with the expectation that the G708N mutation increases homo-oligomerization of the β3 transmembrane helix, it is conceivable that it induces αIIbβ3 clustering via an association of β3 with other membrane proteins. Thus, we characterized the oligomerization of two synthetic 28-residue peptides corresponding to the transmembrane helices of wild-type β3 and the G708N β3 mutant. On SDS-PAGE (SDS–polyacrylamide gel electrophoresis), the mobility of the wild-type β3 helix was consistent with a monomeric transmembrane helical standard, whereas the mobility of the G708N peptide was consistent with a trimer (Fig. 3B).

Equilibrium sedimentation for both peptides fit best with a monomer-trimer equilibrium (fig. S2). When the data were plotted as the relative proportion of different species as a function of peptide/detergent molar ratio (Fig. 3C), the G708N peptide showed a dramatic increase in the fraction of trimer compared with the wild-type peptide. Calculated Kd for the G708N (pKd = 5.24 in mole fraction units) and the wild-type (pKd = 3.42) peptides revealed that the G708N mutation increased the stability of the trimer by more than an order of magnitude.

It was shown previously that integrins have a tendency to associate in vitro (25, 26), and that this association is mediated, at least in part, by their transmembrane domains (11, 27). Consistent with these observations, transmission electron micrographs of purified wild-type αIIbβ3 activated by 1 mM Mn2+ revealed the formation of αIIbβ3 dimers and trimers via an interaction that exclusively involved the distal ends of the transmembrane stalks (Fig. 4A). Moreover, the trimers are open, as would be predicted if the stalks undergo homomeric interactions.

Fig. 4.

(A) Transmission electron microscopy of purified αIIbβ3. Dimers and trimers of αIIbβ3 were observed when the integrin was incubated with activating metal ion Mn2+, but not when the integrin was incubated with Ca2+ (22). (B) A potential mechanism for the activation and clustering of integrin αIIbβ3. In the inactive state the transmembrane and cytoplasmic domains of αIIb (blue) and β3 (red) subunits are proximal. Concomitant with activation, the transmembrane domains become separated and available for homomeric interactions. Following Takagi et al. (6, 32), we picture the inactive conformation as being bent, although this is not an essential feature of our model. Homomeric association of the transmembrane domains leads to αIIbβ3 clustering on the cell surface. In the clustered state, the extracellular domain of each heterodimeric integrin remains in the active conformation. Integrin interacting partners (not included in the figure here), such as extracellular ligands, cytosolic proteins, and cytoskeleton, may regulate integrin activity by affecting the equilibrium among these states.

Here we show that homo-oligomerization has important functional consequences as well. Two different Asn mutations, separated by 10 Å along the β3 transmembrane helix, induced homo-oligomerization in vitro and constitutive integrin activation and clustering in vivo. We have also considered the possibility that the M701N and G708N mutations help activate the integrin by disrupting a heterodimeric association between the αIIb and β3 transmembrane helices in the inactive state of the integrin. However, we find this possibility unlikely. Recently reported nuclear magnetic resonance structural data indicate that the αIIb and β3 cytoplasmic tails are able to physically interact (10). However, extrapolation of the transmembrane helices from this structure (10) shows that they are too far apart to allow inter-helical contacts involving residues 701 and 708. On the basis of this model and others, the α and β transmembrane helices are proposed to interact (28, 29). However, several of the Asn mutants that might have been expected to affect heterodimerization had no effect on αIIbβ3 activity.

αIIbβ3 in platelets exists in either of two affinity states whose relative proportion is determined by platelet stimulation (30). Here, we have demonstrated that the equilibrium between these states can be shifted by enhancing the tendency of the β3 transmembrane domain to undergo homo-oligomerization. Thus, the transmembrane helix-cytoplasmic domain of β3 is appropriately poised to allow dynamic changes in the αIIbβ3 activation state. Moreover, interactions that occur only in the activated state would be expected to stabilize this conformation of αIIbβ3. Thus, homo-oligomerization provides a mechanism for driving the equilibrium toward an activated state, while simultaneously inducing the formation of αIIbβ3 clusters (Fig. 4B). Additional interactions involving the αIIb and/or β3 cytoplasmic domains could finely modulate the overall activation process.

Supporting Online Material

Materials and Methods

Figs. S1 and S2


References and Notes

View Abstract

Navigate This Article