Bidirectional Transmembrane Signaling by Cytoplasmic Domain Separation in Integrins

See allHide authors and affiliations

Science  19 Sep 2003:
Vol. 301, Issue 5640, pp. 1720-1725
DOI: 10.1126/science.1084174


Although critical for development, immunity, wound healing, and metastasis, integrins represent one of the few classes of plasma membrane receptors for which the basic signaling mechanism remains a mystery. We investigated cytoplasmic conformational changes in the integrin LFA-1 (αLβ2) in living cells by measuring fluorescence resonance energy transfer between cyan fluorescent protein–fused and yellow fluorescent protein–fused αL and β2 cytoplasmic domains. In the resting state these domains were close to each other, but underwent significant spatial separation upon either intracellular activation of integrin adhesiveness (inside-out signaling) or ligand binding (outside-in signaling). Thus, bidirectional integrin signaling is accomplished by coupling extracellular conformational changes to an unclasping and separation of the α and β cytoplasmic domains, a distinctive mechanism for transmitting information across the plasma membrane.

Signaling across the cell membrane is typically accomplished by either conformational changes in receptor transmembrane domains, as in G protein–coupled receptors, or by lateral association of two or more monomers, as in receptor tyrosine kinases. Integrins represent a large family of heterodimeric receptors that have the ability to mediate bidirectional signaling. Stimuli received by cell surface receptors, including G protein–coupled chemokine receptors and tyrosine kinase–coupled T cell receptors, initiate intracellular signals that impinge on integrin cytoplasmic domains and trigger increased adhesiveness in the extracellular domain (inside-out priming/signaling). Ligand binding by integrin extracellular domains is transduced to the cytoplasm in the classical outside-in direction to regulate intracellular signals that affect cellular growth, differentiation, and apoptosis (outside-in activation/signaling). However, the basic mechanism by which the relatively small cytoplasmic domains, which themselves lack enzymatic activity, function to receive and transmit intracellular signals remains unknown.

Activation of cell surface integrins is associated with conformational changes in the extracellular domains, as evidenced by the many monoclonal antibodies (mAbs) identified that bind preferentially to the activated and/or ligand-occupied forms of integrins (1). Recent nuclear magnetic resonance (NMR) (2) and electron microscopic (EM) studies (3) of purified extracellular domains have demonstrated large-scale rearrangements associated with integrin activation and ligand binding. In addition, an engineered conformational change in the ligand binding inserted (I) domain of LFA-1 (αLβ2) is sufficient to drive a 10,000-fold increase in affinity for the ligand ICAM-1 (4, 5). Mutation of the membrane proximal cytoplasmic regions of the α and β subunits induces constitutive activation (6, 7), suggesting that cytoplasmic interactions between the α and β subunits regulate the affinity of integrins. Consistent with this explanation, introduction of an artificial clasp between the α and β cytoplasmic domains constrains an integrin to be inactive, whereas release of the clasp induces constitutive activation (8). One recent NMR study suggests a specific interaction that is disrupted by mutations and by binding of the talin head domain (9), whereas another NMR study suggested a different structure for the complex (10), and two other NMR studies have failed to detect interactions (11, 12). An EM study is consistent with association between the transmembrane domains in the detergent solubilized resting αIIbβ3 (13). Thus, although specific and regulated interactions between α and β subunit cytoplasmic domains have been widely speculated to be important for integrin activation, they have never been directly demonstrated in vivo. Here, we use fluorescence resonance energy transfer (FRET) (Fig. 1A) to investigate the spatial proximity of αL and β2 cytoplasmic domains in living cells. These studies provide evidence for conformational changes in the cytoplasmic domains of integrins during physiologic activation.

Fig. 1.

Development of the integrin cytoplasmic domain FRET assay. (A) Schematic representation of integrin activation. In the bent conformation (left), extensive interactions are present between the integrin headpiece and tailpiece, and the juxtamembrane, C-terminal segments of the α and β subunit extracellular domains are in proximity (2, 3, 30). Proximity between CFP and YFP fused to the C-termini of the cytoplasmic domains would result in efficient FRET. In the extended integrin conformation (right), there is no interaction between the α and β subunits in the tailpiece (2, 3). Movement apart of >100 Å of mCFP and mYFP would abolish FRET. Ribbon diagrams are based on crystal and negative stain EM studies of integrin αvβ3 (3, 30) (red and blue), and crystal studies on integrin I domains(1) (green) and GFP (31) (cyan for CFP and yellow for YFP). The orientation between these units is arbitrary. (B and C). Intersubunit FRET measurements. (B) Images of CFP and YFP fluorescence for individual cells before and after YFP bleaching are shown in the upper two rows. Note disappearance of YFP fluorescence after bleaching, the marked increase in CFP fluorescence for m5/m6, and little increase for 26/12. In the lower row, FRET efficiency from 0 (black) to 0.4 (red) is shown for each pixel in the region of interest on the plasma membrane of the cell. Pixel intensity histograms are for each CFP pixel acquired before (red line) and after (blue line) photobleaching the YFP acceptor. (C) Basal intersubunit FRET efficiency. *, P < 0.05 versus 26/12 (n = 7 to 12 cells with acceptor intensity of 250 to 400). Bars show SEM. (D) Interheterodimer FRET between neighboring αL2 molecules for individual cells (triangles) was fit to the saturable one-site binding model E% = E%maxF/(F + K), where FRET efficiency (E%) is a hyperbolic function of the YFP acceptor intensity (F), and K is analogous to a dissociation constant (14). The nonlinear least squares regression fits of FRET efficiency between αL-mCFP/β2 and αL-mYFP/β2 (red curve) yielded K = 700, showing little association. By contrast, cross-linking with TS2/4 mAb and anti-IgG for 30 min at 37° C yielded K = 10.4, showing association between heterodimers. The grey region shows the range of acceptor intensity of cells that were selected for all subsequent intersubunit FRET measurements and the corresponding background values expected for interheterodimer FRET.

In initial fusion constructs, the length of the linker between the C-terminus of αL or β2 and the N-terminus of cyan fluorescent protein (CFP) or yellow fluorescent protein (YFP) was varied. Combinations of these constructs with 26- or 5-residue α subunit linkers and 12- or 6-residue β subunit linkers were transiently transfected into K562 cells, where they were distributed predominantly to the plasma membrane (Fig. 1B). Basal FRET efficiency was measured by the acceptor-photobleach method in which the donor (CFP) fluorescence intensity before and after specific photodestruction of acceptor (YFP) was compared (Fig. 1B). Gradual reduction in the total length of the α and β subunit linkers was associated with a steady increase in FRET (Fig. 1C), suggesting that the cytoplasmic domains are basally in proximity. Constructs containing linkers between αL and CFP and between β2 and YFP of 5 and 6 residues, respectively, exhibited the highest FRET efficiency. Mutations were then introduced into CFP and YFP that inhibit their inherent tendency to form hetero- or homodimers, generating “monomeric” CFP (mCFP) and YFP (mYFP) (14). Basal FRET efficiency for αL-CFP/β2-YFP and αL-mCFP/β2-mYFP heterodimers was similar (5/6 compared with m5/m6, Fig. 1C), and the αL-mCFP and β2-mYFP constructs containing 5- and 6-residue linkers were used for most subsequent experiments.

To address the potential contribution of interheterodimer energy transfer to our measurement of intersubunit FRET, we transiently transfected cells with αL-mYFP and αL-mCFP, together with wild-type β2, and measured energy transfer between αL-mYFP/β2 and αL-mCFP/β2 over a range of cell surface densities as determined by acceptor intensities. Interheterodimer energy transfer increased with increasing integrin density at the cell surface (Fig. 1D). Fitting the data to a previously described equation (Fig. 1D, red line) (14) showed little tendency of αLβ2 heterodimers to associate or cluster in the membrane. At relatively low density (acceptor intensities from 250 to 400 arbitrary fluorescence units), energy transfer efficiency between αL-mYFP/β2 and αL-mCFP/β2 was ∼0.07. However, FRET efficiency between αL-CFP and β2-YFP (5/6) or αL-mCFP and β2-mYFP (m5/m6), measured in the same range of acceptor intensity between 250 and 400, was 0.17 to 0.2 (Fig. 1C), demonstrating that this energy transfer occurs primarily through intersubunit FRET within individual heterodimers. All subsequent FRET measurements were made on cells with acceptor intensities of 250 to 400.

Stable K562 cell transfectants expressing αL-mCFP/β2-mYFP and αL-CFP/β2-YFP were generated, and clones expressing acceptor intensity ranging from 250 to 400 units were selected. Most of the LFA-1 in these cells, as with the transient transfectants, was distributed to the plasma membrane (Fig. 2A). Some Golgi fluorescence was also evident, but was omitted by measuring FRET only in the rim around each cell. SDS–polyacrylamide gel electro-phoresis (SDS-PAGE) of cell lysates of the stable transfectants and Western blotting demonstrated that both αL-CFP and β2-YFP were increased in Mr compared with wild-type αL and β2 (Fig. 2B, left and center) by an amount corresponding to the Mr of the fused CFP or YFP. αL-CFP and β2-YFP were also recognized by anti–green fluorescent protein (GFP), which cross-reacts with CFP and YFP (Fig. 2B, right). No evidence of proteolytic cleavage of CFP or YFP was detected (Fig. 2B, right). Clones expressing similar levels of cell surface αL-CFP/β2-YFP, αL-mCFP/β2-mYFP, and wild-type αL2 were selected (Fig. 2C).

Fig. 2.

Stable cells expressing αL-mCFP/β2-mYFP preserve normal functional profiles of LFA-1. (A) Fluorescence images of stably transfected K562 cells with αL-CFP/β2-YFP or αL-mCFP/β2-mYFP demonstrate membrane localization of CFP and YFP signals. (B) Whole cell lysates of K562 and stable transfectants expressing wild-type αL2 or αL-CFP/β2-YFP were subjected to SDS-PAGE and Western blotting with the indicated antibodies. (C) Cell surface expression levels of αL2 were determined by immunofluorescence flow cytometry on parental K562 cells or stable transfectants. mAbs TS1/22, CBR LFA-1/7, and TS2/4 recognize αL, β2, and αL2 complex, respectively. Nonbinding X63 IgG1 was used as negative control. (D) Adhesion to immobilized ICAM-1. K562 transfectants were allowed to bind ICAM-1–coated substrates without stimulation [control (Con)] or after stimulation with 1 mM Mn2+, 10 μg/ml CBR LFA-1/2 mAb, or 100 nM PMA. Data are mean ± SEM for one representative experiment in triplicate. (E) KIM127 activation epitope induction by Mn2+ or PMA.

The fusion constructs were tested for functional responses. The wild-type αL2, αL-mCFP/β2-mYFP, and αL-CFP/β2-YFP transfectants showed low basal adhesion to the immobilized ligand ICAM-1 (Fig. 2D) and low basal staining with KIM127 (Fig. 2E), a mAb that binds near the genuflection in the β2 subunit and reports the active, extended conformation of LFA-1 (2, 15, 16). In cells expressing wild-type αL2 and αL-mCFP/β2-mYFP, ICAM-1 binding and KIM127 staining were greatly increased by Mn2+ (17, 18), CBR LFA-1/2, an activating mAb to β2 (15), and phorbol 12-myristate 13-acetate (PMA) (19) (Fig. 2D and E). By contrast, cells expressing αL-CFP/β2-YFP, containing the dimerizing form of CFP and YFP, bound less efficiently to ICAM-1 when activated by Mn2+ and CBR LFA-1/2, and binding could not be activated by PMA (Fig. 2D). In addition, PMA was less effective than Mn2+ in enhancing KIM127 staining in αL-CFP/β2-YFP (Fig. 2E).

To determine whether integrin activation is associated with spatial rearrangements of the α and β subunit cytoplasmic tails in living cells, FRET efficiency was measured on cells expressing αL-mCFP/β2-mYFP and αL-CFP/β2-YFP. In cells expressing αL-mCFP/β2-mYFP, FRET was significantly decreased from basal levels after treatment with CBR LFA-1/2 mAb, Mn2+ + CBR LFA-1/2 mAb, or PMA (Fig. 3A and C). However, no significant clustering or redistribution of LFA-1 in the plasma membrane (20, 21, 22) was induced by the activating treatments used in this study (Fig. 3C); furthermore, clustering would have increased rather than decreased FRET efficiency. These results show that activation mediated by CBR LFA-1/2 mAb and PMA is associated with spatial separation of the αL and β2 cytoplasmic domains. In contrast, in cells expressing αL-CFP/β2-YFP, CBR LFA-1/2 mAb and PMA did not alter FRET efficiency (Fig. 3B). These results are explained by the previous observations that CFP and YFP, but not mCFP and mYFP, mediate dimerization when membrane-associated (14), and that an association enforced by substitution of integrin α and β subunit cytoplasmic domains with α-helical coiled-coil domains blocks activation (8). Thus, the results suggest that in αL-CFP/β2-YFP, CFP and YFP form a heterodimer that opposes cytoplasmic domain separation and perturbs normal integrin activation. By contrast, αL-mCFP/β2-mYFP functions indistinguishably from wild-type αLβ2. Mn2+ had no effect on FRET in either αL-mCFP/β2-mYFP or αL-CFP/β2-YFP transfectants (Fig. 3, A and B), in contrast with its activation of KIM127 epitope exposure and adhesion (Fig. 2E). This result suggests that Mn2+-mediated activation can take place independently of αL and β2 cytoplasmic domain separation.

Fig. 3.

Activation of αL-mCFP/β2-mYFP causes spatial separation of the cytoplasmic domains. (A and B) FRET was measured in αL-mCFP/β2-mYFP (A) or αL-CFP/β2-YFP (B) K562 transfectants after the indicated treatment. Data are averages ± SEM for 5 to 10 cells. *, P < 0.05 versus control (Con). (C) Representative single-cell FRET measurements. Note the greater post-YFP bleach increase in CFP fluorescence for the control cell. Data are displayed as in Fig. 1B, except only CFP fluorescence is shown. (D) Effect of GFFKR motif mutation on FRET. Data show mean ± SEM for 4 to 10 cells. *, P < 0.05 versus αL-mCFP/β2-mYFP.

Binding to ligands by integrins transduces signals into cells. Electron micrographic studies of the extracellular fragment of αVβ3 integrin show that binding to ligand-mimetic cyclic Arg-Gly-Asp (RGD) peptides shifts the conformational equilibrium completely toward the extended form with the open headpiece, which has widely separated C-terminal juxtamembrane domains, whereas Mn2+ shifts the equilibrium only partially, so that the bent conformation and extended conformations with open and closed head-pieces coexist (3). The presence of transmembrane and cytoplasmic domains would affect the conformational equilibrium. To examine the effect of ligand binding on cytoplasmic domain association, we incubated cells with soluble ICAM-1 in the presence of Mn2+, which increases the affinity of LFA-1 sufficiently to obtain appreciable binding to soluble ICAM-1 (17), and compared the results with results in Mn2+ alone. Binding of soluble ICAM-1 to αL-mCFP/β2-mYFP in the presence of Mn2+ induced a significant increase in KIM127 staining compared with Mn2+ alone (not shown) and a dramatic decrease in FRET efficiency compared with cells treated with Mn2+ alone (Fig. 3A). This result shows that ligand binding induces spatial separation of the cytoplasmic domains and defines a key step in outside-in signal transduction.

Mutation of the conserved GFFKR sequence at the boundary between the integrin α subunit transmembrane and cytoplasmic domains results in constitutive adhesiveness (6, 7) as well as increased exposure of extracellular domain activation epitopes (7), suggesting that the GFFKR sequence stabilizes the inactive integrin conformation. Thus, the GFFKR motif of αL-mCFP was mutated to GAAKR, and K562 cells were transiently transfected with this construct and β2-mYFP. FRET efficiency measured under basal conditions was significantly decreased compared with αL-mCFP/β2-mYFP transfectants (Fig. 3D). Thus, constitutive activation induced by perturbation of the α subunit GFFKR sequence is associated with basal spatial separation of the α and β subunit cytoplasmic domains.

Talin is a 250-kD cytoskeletal protein that is composed of a 47-kD N-terminal head domain and a 190-kD C-terminal rod domain. When freed from association with the rod domain by proteolysis or truncation, the talin head domain activates integrin αIIbβ3 (9, 23, 24). Talin binds directly to β1, β2, and β3 integrins (25), but whether it activates β2 integrins has not been examined. αL-mCFP/β2-mYFP K562 transfectants were further transfected with hemagglutinin (HA) tagged talin head domain (TH), and clones expressing no (TH null), medium (TH middle), or high (TH high) levels of talin head domain were selected (Fig. 4A). Expression of the talin head domain did not change the surface expression or distribution pattern of αL-mCFP/β2-mYFP. Immunoprecipitation with CBR LFA-1/2 mAb to the β2 subunit followed by immunoblotting with HA antibody showed that the talin head domain associated with LFA-1 (Fig. 4B). Basal adhesiveness to immobilized ICAM-1 was significantly enhanced in cells expressing middle and high levels of talin head domain (Fig. 4C), demonstrating that the talin head domain activates αLβ2, as previously shown for αIIbβ3 (9, 23, 24). In contrast, TH null cells behaved similarly to parental αL-mCFP/β2-mYFP cells (Fig. 4C). Basal FRET efficiency in talin head domain transfectants revealed significant decreases in energy transfer between αL-mCFP and β2-mYFP, which were proportional to the level of talin head domain expression (Fig. 4D). These data demonstrate that talin head domain binding to αLβ2 causes a spatial separation of the cytoplasmic domains, consistent with NMR studies on dissociation, by talin head domain, of αIIbβ3 cytoplasmic complex (9).

Fig. 4.

Effect of the talin head domain on integrin LFA-1 cytoplasmic domain association and adhesion to ICAM-1. All experiments are on cells expressing αL-mCFP/β2-mYFP and different amounts of the N-terminally HA-tagged talin head domain. (A) Talin head domain expression. Cells were subjected to SDS-PAGE and Western blotting. Blots were probed with antibody to HA (upper), then stripped and reprobed with a mAb specific for the talin rod domain. (B) Binding of the talin head domain to LFA-1. Triton X-100 lysates were immunoprecipitated with CBR LFA-1/2 mAb to β2 or X63 IgG1 as negative control, and subjected to SDS-PAGE and Western blotting with antibody to HA (upper), then stripped and reprobed with GFP antibody (lower). (C) Adhesion of transfectants to ICAM-1 substrates was measured in the presence of Mn2+ or CBR LFA-1/2 mAb. Data are mean ± SEM for three to six experiments, each in triplicate. *, P < 0.05 versus parental control. (D) FRET in talin head domain transfectants. Each square represents FRET efficiency of one cell. Bar = mean. *, P < 0.05 versus parental control.

Physiologic regulation of integrins can occur through a variety of cell surface receptors. One of the best characterized examples is activation of integrin-dependent leukocyte arrest and migration through G protein–coupled receptors for chemokines (2628). To assess the effect of chemokine receptor signaling on the conformation of integrin cytoplasmic domains, αL-mCFP/β2-mYFP transfectants were further transfected transiently with HA-tagged CXCR4. CXCR4 is the receptor for the chemokine SDF-1. Both cells that were positive and cells that were negative for CXCR4 expression were identified by immunofluorescence with an antibody to HA and Cy5-labeled anti-Ig (Fig. 5A). Addition of the chemokine SDF-1 significantly reduced FRET between αL-mCFP and β2-mYFP in cells expressing CXCR4 but had no effect on cells lacking CXCR4 expression (Fig. 5B). Furthermore, treatment with 12G5, a blocking antibody to CXCR4, abolished the effect of SDF-1 on FRET efficiency (Fig. 5B). This result demonstrates that physiological “inside-out” activation of integrins stimulated by chemokine receptors is associated with spatial separation of the α and β cytoplasmic domains of LFA-1.

Fig. 5.

Effect of chemokine-receptor activation on LFA-1 cytoplasmic domain FRET. (A) Cells expressing αL-mCFP/β2-mYFP were transiently transfected with N-terminally HA-tagged CXCR4 receptor. Transfectants were identified by staining with mouse antibody to HA and Cy5-anti-mouse Ig (arrow). (B) Effect of SDF-1 treatment on FRET. Transfectants with or without pretreatment with 10 μg/ml 12G5 mAb were treated with or without 1 μg/ml SDF-1α. FRET was measured both on transfectants that expressed (+) or did not express (–) CXCR4, as shown with Cy5 fluorescence. *, P < 0.05 versus untreated CXCR4 transfectants. Data are mean ± SEM for four cells.

Integrins are exceptional among regulated cell surface receptors in that the large size of their extracellular domains requires that conformational information travel great distances to couple extracellular ligand binding to the intracellular cytoplasmic domains. The problem of how to transmit information between the membrane and the ligand binding site, which are >200 Å apart in the extended conformation, is in part solved by equilibration with a bent conformer, which has an αβ interface close to the membrane (3). However, the mechanism by which information flows bidirectionally across the plasma membrane in integrins has remained a mystery. The results presented here demonstrate conformational changes in integrin cytoplasmic domains associated with activation in living cells. Because many of the stimuli studied here reduced αL2 FRET to a level similar to that of the measured background interheterodimer αLL FRET, and interheterodimer FRET will still be present after maximal αL2 separation, separation appears to approach the detection limit for CFP-YFP FRET of ∼100 Å. Orientation is highly unlikely to influence FRET between αL-mCFP and β2-YFP because of the flexibility of the 5- and 6-residue GPVAT and GGPVAT linkers. Our studies with probes attached to the C-termini of the short integrin αL and β2 cytoplasmic domains do not define the nature of coupled movements in the αL and β2 transmembrane domains; nonetheless, the large changes in FRET that are observed appear more consistent with the most dramatic model of alterations in the transmembrane domain association, i.e., movement apart in the plane of the membrane with complete separation of the transmembrane domain (3, 8, 29), rather than with more subtle models of hinging or pistonlike motions.

We conclude that integrin priming stemming from chemokine receptor–stimulated inside-out signaling, talin head domain binding, PMA, or GFFKR mutation is associated with a significant spatial separation of the α and β subunit cytoplasmic domains (Fig. 1A). Furthermore, outside-in signals transmitted after ligand binding also induce separation of the cytoplasmic domains. Thus, in integrin signaling across the plasma membrane, spatial separation of cytoplasmic domains is coupled to extracellular conformational change, and these mechanisms operate in both the inside-out and outside-in directions. This result demonstrates a distinctive mechanism for transmission of information across the plasma membrane by cell surface receptors.

Supporting Online Material

Materials and Methods


References and Notes

View Abstract

Navigate This Article