Native Ligands Change Integrin Sequestering but Not Oligomerization in Raft-Mimicking Lipid Mixtures

Abstract

Distinct lipid environments, including lipid rafts, are increasingly recognized as a crucial factor affecting membrane protein function in plasma membranes. Unfortunately, an understanding of their role in membrane protein activation and oligomerization has remained elusive due to the challenge of characterizing these often small and transient plasma membrane heterogeneities in live cells. To address this difficulty, we present an experimental model membrane platform based on polymer-supported lipid bilayers containing stable raft-mimicking domains (type I) and homogeneous cholesterol-lipid mixtures (type II) into which transmembrane proteins are incorporated (α_vβ₃ and α₅β₁ integrins). These flexible lipid platforms enable the use of confocal fluorescence spectroscopy, including the photon counting histogram method, in tandem with epifluorescence microscopy to quantitatively probe the effect of the binding of native ligands from the extracellular matrix ligands (vitronectin and fibronectin for α_vβ₃ and α₅β₁, respectively) on domain-specific protein sequestration and on protein oligomerization state. We found that both α_vβ₃ and α₅β₁ sequester preferentially to nonraft domains in the absence of extracellular matrix ligands, but upon ligand addition, α_vβ₃ sequesters strongly into raft-like domains and α₅β₁ loses preference for either raft-like or nonraft-like domains. A corresponding photon counting histogram analysis showed that integrins exist predominantly in a monomeric state. No change was detected in oligomerization state upon ligand binding in either type I or type II bilayers, but a moderate increase in oligomerization state was observed for increasing concentrations of cholesterol. The combined findings suggest a mechanism in which changes in integrin sequestering are caused by ligand-induced changes in integrin conformation and/or dynamics that affect integrin-lipid interactions without altering the integrin oligomerization state.

Introduction

The modern view of the plasma membrane is that it is a highly complex fluid system comprised of rapidly changing heterogeneous patches of lipids that may regulate the location and functionality of membrane-associated proteins (1,2). One such class of heterogeneous patches consists of lipid rafts, which are defined as dynamic assemblies enriched in cholesterol (CHOL), sphingolipids, and GPI-anchored proteins (3). Rafts of membrane proteins, in concert with other proteins or lipids, induce a variety of cellular activities, including formation of signaling platforms (4), pathogenesis through endocytosis (5), and changes in cellular adhesion, cell morphology, and angiogenesis (6). An important aspect of lipid rafts is that protein functionality can be regulated by a change in raft association or sequestration (7).

Several possible mechanisms may induce a change in sequestration of receptors. Sequestration of proteins into lipid rafts may arise due to protein acylation (8), receptor clustering (9), ligand addition (10), or other specific protein-protein interactions (11). CHOL concentration is known to be a critical factor in sequestration events, as exemplified by CHOL depletion studies that showed a change in protein activity level that was restored when CHOL returned to normal levels (12,13). However, the underlying mechanisms of protein sequestering, and the interplay among them, are still not well understood because it can be very challenging to distinguish contributions from different factors in plasma membranes. CHOL depletion, for example, may induce artifacts such as cytoskeletal destabilization (14). Furthermore, the common practice of identifying proteins found in detergent-resistant membranes with rafts appears to be, in part, prone to inaccurate analyses (15). Rafts are small in size, and protein-lipid raft associations are dynamic and often short-lived (16,17). One method for increasing raft size and duration in live cells is cross-linking, i.e., through cross-linking antibodies, GM1-cholera toxin B (CTxB) cross-linking, or mechanical cross-linking of ligands (18–21). These methods, however, artificially induce cell responses. Lingwood and Simons (22) called this problem, akin to Heisenberg's uncertainty principle, the observer effect of lipid raft studies: before heterogeneity can be observed in plasma membranes, it must first be induced. Consequently, model studies containing raft-mimicking lipid mixtures have also been pursued. In silico progress has been slow, in part due to the limited availability of complete crystal structures for transmembrane proteins. Synthetic bilayers provide a promising avenue of investigation. If lipid bilayers are composed of ternary mixtures of CHOL, saturated lipids, and unsaturated lipids, they can phase-separate into regions of higher order with more CHOL (l_o) and lower order with less CHOL (l_d) (23,24). Sequestration of membrane receptors and response of sequestration to cross-linking antibodies have also been confirmed in model membrane environments (25–27).

The elegance of model systems lies in the fact that they enable one to study the sequestering and oligomerization behaviors of receptors in response to biological stimuli without the use of artificial cross-linking agents. Applying this concept, we expanded the model membrane approach to study the effects of ligand addition to the integrins α_vβ₃ and α₅β₁ incorporated into ternary phase-separating planar polymer-tethered model lipid bilayers. The polymer-tethered planar system enables stable lipid phase separations (28,29), incorporation of biologically active transmembrane proteins (30–32), and sensitive fluorescence fluctuation spectroscopy investigations into lipid phase partitioning and protein oligomerization state (33). Integrins are well suited for these experiments because they are implicated in many raft-associated activities, including cell adhesion, morphology, motility, and angiogenesis. They also act as bidirectional signaling platforms in which functionality is regulated by different factors, including binding of ligands or other membrane proteins, activation through divalent cations, signals from the cytosolic environment, and microclustering (6). Using confocal spectroscopy XY scans, we show that α_vβ₃ and α₅β₁ in a polymer-tethered lipid bilayer in PBS buffer partition preferentially to l_d domains, and ligand addition alone causes substantial, quantifiable shifts in protein partitioning to l_o domains. Despite these substantial shifts, ligand addition had essentially no impact on receptor oligomerization state, as determined by photon counting histograms (PCHs) for receptors incorporated into binary lipid/CHOL bilayers or ternary raft-mimicking bilayers. By contrast, independently of ligand addition, the concentration of CHOL in binary bilayers affected receptor oligomerization to a moderate degree.

Materials and Methods

Materials

The lipopolymer 1,2-dioctadecyl-sn-glycero-3-N-poly(2-methyl-2-oxazoline)₅₀ (diC₁₈M₅₀) was synthesized according to previously described procedures (34). Lipopolymer-coated quantum dots were synthesized according to established procedures (35). The phospholipids 1-stearoyl-2-oleoyl-sn-glycero-3-phosphocholine (SOPC), 1,2-dioleoyl-sn-glycro-3-phosphocholine (DOPC), 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC), 1,2-dipalmitoyl-sn-glycero-3-phosphothioethanol (DPTE), 1,2-dioleoyl-sn-glycro-3-phosphoethanolamine (DOPE), CHOL,) and ganglioside GM1 (ovine brain) were purchased from Avanti Polar Lipids (Alabaster, AL). The fluorescently labeled phospholipids N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)-1,2-dihexadec-anoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt (NBD-DHPE), N-(6-tetramethylrhodamine-thiocarbamoyl)-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt (TRITC-DHPE), Alexa-555 labeled CTxB (CTxB-555), and kits for fluorescently labeling antibodies with TRITC or Alexa-555 were obtained from Invitrogen (Carlsbad, CA). Chloroform (HPLC grade; Fisher Scientific, Pittsburgh, PA) was used as a spreading solvent for lipid monolayers at the air-water interface, and Milli-Q water (pH = 5.5, 18 MΩ-cm resistivity; Millipore, Billerica, MA) was used as a subphase material for all experiments. Glass coverslips were pretreated by baking for 3 h at 515°C, followed by sonication in a bath sonicator for 45 min first in 1% SDS, then NaOH saturated MeOH, and finally 0.1% HCl (Fisher Scientific), with extensive rinsing between sonication steps and after the final sonication step, and stored in Milli-Q for no more than 1 week. The proteins human integrin α_vβ₃ and α₅β₁, octyl-β-D-glucopyranoside (OG) formulation; monoclonal antibodies (MAbs) anti-integrin α_vβ₃, clone LM609, and anti-integrin α₅β₁, clone JBS5; vitronectin, human purified (VN); and fibronectin, human purified (FN) were purchased from Millipore (Billerica, MA). Rhodamine 6G was purchased from Sigma-Aldrich (St. Louis, MO). Antibodies were labeled as described in the antibody labeling kits. Efficacy of labeling was checked by determining brightness by fluorescence correlation spectroscopy (FCS).

Construction of polymer-tethered phospholipid bilayers

Polymer-tethered phospholipid bilayers were built with the use of successive Langmuir-Blodgett (LB) and Langmuir-Schaefer (LS) film transfers according to standard procedures (28). To form the first (LB) monolayer, a chloroform solution of a mixture of diC₁₈M₅₀ and lipids was spread at the air-water interface of a film balance with dipper (Labcon, Darlington, UK). The composition was 5 mol % diC₁₈M₅₀, 31.5 mol % DPPC and CHOL, and 32% DOPC for type I bilayers. For type II bilayers, the composition consisted of 5 mol % diC₁₈M₅₀ and 0, 5, or 30 mol % CHOL, with the balance consisting of SOPC. The monolayer was compressed and, after stabilization (30 min for type I bilayers, 20 min for other mixtures), transferred to a glass substrate (28). Next, a chloroform solution containing a lipid mixture (the LS mixture) was spread at the air-water interface and compressed. For type I bilayers, the LS mixture consisted of 1:1:1 DPPC/CHOL/DOPC. For type II bilayers, the LS mixture consisted of 0, 5, or 30 mol % CHOL, with the balance consisting of SOPC. LS transfer was accomplished by stabilizing the LS monolayer with a depression slide positioned underneath the air-water interface, and then carefully pushing the glass substrate containing the LB layer onto the underlying depression slide. With the use of a transfer dish, the depression slide was removed and the bilayer was transferred into a petri dish, where the Milli-Q was replaced by PBS (Fisher Scientific, 10× concentration, diluted in Milli-Q).

Incorporation of proteins into bilayers

Proteins were reconstituted into model bilayers by means of a modified Rigaud technique (36), also known as the direct protein incorporation method (37). All protein studies were carried out in PBS buffer at room temperature. Membrane proteins stabilized in 100 mM OG (1.3 × 10⁻¹¹ mol proteins leading to bilayer concentrations of at most 10⁻³ mol %) were added to the PBS solution above the bilayer (6 mL of PBS) for 1.5–2 h, followed by removal of surfactant using a single layer of SM-2 biobeads previously slurried in PBS (Bio-Rad, Hercules, CA) applied for 15 min to solution over bilayers. Incorporation of proteins into type I bilayers was assisted by initially diluting the proteins into 1 mL of 250 μM OG before addition to the PBS solution above the bilayer, leading to a detergent concentration in the presence of the bilayer of ∼0.002 cmc, but this step was not necessary for type II bilayers. Presoftening of the bilayer by addition of OG or other detergent before protein incorporation (37–39) was not required at these low protein concentrations (250 pM) and long incubation times. Extensive rinsing in PBS ensured efficient removal of surfactant, as confirmed by analysis of domain shapes (see Fig. S1 in the Supporting Material) and lipid fluidity as discussed below. Fluorescently labeled MAbs were then added for 2–4 h and excess antibodies were removed by rinsing. Antibody binding assays were used to confirm the functional reconstitution of integrins, and fluidity was also checked. After data were acquired on these systems, FN or VN was added in a 1:1 ratio to membrane proteins and permitted to equilibrate 3–4 h, and data were taken on the same substrates. Several experiments were conducted at 12 h after ligand addition, and no differences were observed. Unbound FN and VN were removed by rinsing before imaging. MAbs were added to bilayers in the absence of integrins as a control to check for nonspecific binding of antibodies.

Microscopy techniques

A commercial ConfoCor 2 (Zeiss, Jena, Germany) was used for epifluorescence (EPI) microscopy and fluorescence fluctuation spectroscopy (FFS). EPI was conducted with the use of an inverted optical microscope (Axiovert 200M, Zeiss, Oberkochen, Germany) in which the beam was focused to the sample by a microscopy objective (C-Apochromat, water immersion, 40× NA = 1.2; Zeiss). EPI studies were undertaken with a Zeiss AxioCam MRm monochrome digital camera and Axiovision 4.8 software or a CoolSNAP_fx camera (Roper Scientific, Princeton, NJ) and Roper Scientific imaging software. FFS data were acquired using a 1.8 mW HeNe laser (543 nm) with a 560–605 nm emission filter (red channel) or a 30 mW Argon laser (514 nm) with a 500–530 emission filter (green channel). We performed confocal spectroscopic XY (CS-XY) scans (10 × 10 μm, 0.5 μm steps) of both channels after taking EPI micrographs of the same location. Control CS-XY scans were taken of bilayers with NBD-PE and no proteins to determine the amount of bleed-through from the green channel to the red channel. FFS data for FCS and PCH analyses were acquired for 50 s runs on the bilayer or in solution, using the same pinhole size, with focusing of planar systems accomplished by maximizing peak count rates from single fluorescent molecules diffusing through the confocal volume. Ternary lipid mixtures (type I) induced phase separations into l_o (liquid-ordered) and l_d (liquid-disordered) domains, which were visualized by EPI through one of three NBD-PE labeling strategies: 1), addition of 0.5 mol % NBD-PE directly to the bilayer with a correction for background bleed-through in the CS-XY scans; 2), addition of 0.2 mol % NBD-PE (sufficiently low to avoid the need for background correction); or 3), addition of NBD-PE only after completion of all protein data through the use of fusogenic vesicles (35). All three strategies provided similarly accurate quantitative analyses of protein distributions. We investigated protein diffusion using wide-field, single-molecule fluorescence microscopy of membrane proteins in dilute (10⁻⁸) concentrations after fluorescent MAb labeling on an inverted microscope (Zeiss Axiovert S100TV) as previously described (40). Further details about this methodology are provided in the Supporting Material. We investigated lipid diffusion for the purpose of analyzing bilayer fluidity by incorporating 2 × 10⁻³ mol % TRITC-DHPE into type I and type II bilayers, and using FCS to determine characteristic diffusion times for the lipids before and after addition of detergent and rinsing with biobeads.

Data analysis

We analyzed integrin sequestering by identifying partition coefficients (K_p (l_o/l_d)), defined as I_lo/I_ld (41) from CS-XY scans. Raw scans were corrected for NBD and background contributions to determine protein signal average intensities in l_o phase (I_lo) and l_d phase (I_ld). To quantify changes in K_p, we introduce E_raft, defined as the difference in signal intensities between l_o and l_d phases normalized by the sum of the signal intensities.

$$E_{raft}=\frac{\left(I_{lo}-I_{ld}\right)}{\left(I_{lo}+I_{ld}\right)}.$$ (1)

For example, for a system perturbed by the addition of ligands, the change in raftophilic excess, divided by two, equals the fraction of proteins X_migrate that have migrated from l_d to l_o:

$$X_{migrate}=\frac{\left(E_{raft\left(+ligand\right)}-E_{raft\left(-ligand\right)}\right)}{2}.$$ (2)

We monitored integrin oligomerization behavior using PCHs of avalanche photodiode photon counts (42,43). To ensure that the data were sufficiently robust above background for the low-concentration diffusing proteins on the bilayer, we also analyzed the intensity trace. We found that when the average signal for the top 1% of the counts collected by intensity trace were at least 20-fold above the average count rate, sufficient data were available to fit the PCH data to a PCH algorithm-generated curve. After finding the average number and brightness of the fluorescent particles (N_avg, ε), as described more fully in the Supporting Material, we refitted the data to look for dimers by assuming a second fluorescent species of twice the brightness and finding the average number for both the primary species and secondary species (N_avg, ε, N_avgdimer, ε_dimer (= 2ε)). N_avg and N_avgdimer were then used to determine X_dimer (mole fraction dimers). We tested the PCH model using fluorescent particles known to be monomers, such as Rhodamine 6G, MAbs in solution, and TRITC-DHPE in a bilayer, and found that these species had X_dimer < 1% (data not shown). The data were also refit with a model encompassing tetramers, which showed correspondingly similar results for both the control systems and the membrane proteins of interest (data not shown). It was also necessary to calibrate the PCH algorithm for brightness of particles in solution relative to particles in a thin slab, such as a planar bilayer (44). This was done by constructing a bilayer with SOPC and 2 × 10⁻² mol % DPTE and adding maleimide functionalized quantum dots (35). We compared the PCH calculations for brightness of the QDs on the bilayer with the brightness of dilute QDs in solution, and determined that the brightness of the species in solution and the brightness of the species on the bilayer were within 5% of each other (Fig. S2).

Results and Discussion

Determination of fluidity of α_vβ₃ and α₅β₁ incorporated into model bilayers

Type I and type II lipid bilayers with 5 mol % DiC₁₈M₅₀ in the LB layer were constructed, and integrins were incorporated therein and subsequently labeled with MAbs as described above. We checked the integrity of the bilayers before protein addition and after detergent extraction by EPI analysis of the bilayers and by determining characteristic diffusion times for lipid probes using FCS. These experiments showed no change in bilayer domain structure (Fig. S1) or lipid fluidity for both type I and type II (0% CHOL) bilayers within the experimental uncertainty of 5% (data not shown). The EPI analysis showed no statistically significant aggregation of integrins on the bilayer. To quantify the functional reconstitution and lateral fluidity of integrins in planar membrane systems, we determined the lateral diffusion of α_vβ₃ and α₅β₁ reconstituted into type II (0% CHOL) bilayers before (−VN/FN) and after (+VN/FN) ligand binding. The calculated diffusion coefficients were as follows: for α_vβ₃, D = 0.39 ± 0.04 (−VN), 0.42 ± 0.04 (+VN); for α₅β₁ D = 0.71 ± 0.07 (−FN), 0.87 ± 0.08 (+FN; all values μm²s⁻¹), immobile fraction (IF) < 10% by inspection of cumulative distribution functions). These diffusion data are in good agreement with data for α_IIbβ₃ integrins reconstituted onto other planar model membranes, but differ in terms of IF (45,46). The differences in IF can be attributed to the different protein labeling and reconstitution strategies used. (In the α_IIbβ₃ system, integrins are fluorescently labeled using 5-(and-6)-carboxy-tetra-methylrhodamine succinimidyl ester and then are reconstituted via proteosomes, in which both membrane protein orientations in the planar membrane system are likely to occur with similar probabilities. In this case, a substantial IF can be expected because integrins pointing with their ectodomains toward the solid substrate are immobile. The DPI reconstitution method, as used in the current study, has been shown to lead to a more unidirectional incorporation of transmembrane proteins (37). By monitoring proteins through the use of fluorescent antibodies that only bind to the protein's extracellular domain, this work also assured that only the properties of correctly oriented proteins (extracellular domain points to water phase) are analyzed, which are less prone to immobilization in the planar bilayer.)

Determination of raft sequestration of proteins before and after ligand binding

To test the sensitivity of the experimental EPI/FFS setup, we constructed type I bilayers with registered l_o and l_d domains with an addition to the LS mixture of 2 × 10⁻³ mol % GM1 and 0.5 mol % of the lipid raft marker NBD-PE. CTxB-555 was subsequently added and CS-XY scans were performed. As expected, the raftophilic CTxB colocalized with the NBD-PE: E_raft for CTxB = 0.68 ± 0.07. This corresponds to a K_p of 5.2 ± 1.2, verifying the sensitivity of the type I bilayers to induce sequestration in the GM1/CTxB system. Next, type I bilayers with registered l_o and l_d domains were constructed with 0.5 mol % NBD-PE added to the LS mixture, and α_vβ₃ was incorporated as described above. After EPI micrographs were taken (Fig. 1, C and F), CS-XY scans were performed, representatives of which are shown in Fig. 1, where the same area is scanned by the red (protein) channel (A and D) and the green (NBD) channel (B and E). A comparison of the CS-XY scans shows that before addition of VN, α_vβ₃ displays a marked preference for the nonraft l_d phase (Fig. 1, A and B). As illustrated in Fig. 1, D–F, adding VN to the same substrate induced a dramatic change in raft preference. The α_vβ₃ proteins switched from an l_d preference to an l_o preference (Fig. 1 D). The NBD-PE distribution was not perturbed by the addition of VN (Fig. 1 F). Motivated by the question of whether this large change in raftophilicity was specific to the α_vβ₃ integrins or was more generally applicable to other integrins, we performed the experiment again, this time using α₅β₁ integrins and FN as the ECM ligand; the results are shown in Fig. 2. In similarity to the α_vβ₃ integrins, before addition of ligand, the α₅β₁ partitioned preferentially to the l_d phase (Fig. 2 A). Subsequent to FN addition, however, the α₅β₁ switched to displaying no strong preference for either the l_d or the l_o phase (Fig. 2 D). The results from the CS-XY scans are quantified and tabulated, along with the partitioning behavior of the GM1-CTxB system, in Fig. 3, which shows E_raft for α_vβ₃ and α₅β₁ before and after ligand binding (VN and FN, respectively), and E_raft for CTxB-555 linked to GM1. The fraction of receptors that translocated from the l_d domains to l_o domains can be quantified as X_migrate, as discussed above (Eq. 2). The CS-XY scans showed X_migrate = 53 ± 6% for α_vβ₃ integrins, and X_migrate = 27 ± 3% for α₅β₁ integrins.

Figure 1.

(A) CS-XY scan of α_vβ₃ − VN. (B) CS-XY scan of NBD-PE in the same location as A. (C) EPI micrograph of NBD-PE, same location as A. (D) CS-XY scan of α_vβ₃ + VN. (E) CS-XY scan of NBD-PE, same location as D. (F) EPI micrograph of NBD-PE, same location as D. (A–F) Box = 10 × 10 μm².

Figure 2.

(A) CS-XY scan of α₅β₁ − FN. (B) CS-XY scan of NBD-PE same location as A. (C) EPI micrograph of NBD-PE in the same location as A. (D) CS-XY scan of α₅β₁ + FN. (E) CS-XY scan of NBD-PE, same location as D. (F) EPI micrograph of NBD-PE, same location as D. (A–F) Box = 10 × 10 μm².

Figure 3.

Normalized difference in intensity between l_o phase and l_d phase (E_raft) shown for GM1-CTxB and α_vβ₃ and α₅β₁ before and after ligand addition (light bars: − ECM ligand; dark bars: + ECM ligand). Negative values of E_raft correspond to K_p < 1.

Cell studies have shown that inactive and unbound integrins, similarly to the proteins reconstituted into the model bilayers in PBS buffer in our experiment, are nonraft-associated (19,48). By contrast, integrin signaling and involvement in cell adhesion, motility, and angiogenesis are all raft-related activities, as demonstrated by studies involving alteration of plasma membrane CHOL levels or association with other raft-associated proteins (49–52). It appears that in the model system presented here, binding to ECM ligands alone suffices to increase a preference in α_vβ₃ and α₅β₁ for the CHOL-rich (l_o) phase, even in the absence of other protein cofactors, activating cations, or known cross-linking agents such as CTxB or cross-linking MAbs. It is notable that α_vβ₃ is more sensitive to raft sequestration than α₅β₁. This finding is intriguing in light of the observation that α_vβ₃ regulates the adhesive and phagocytic activity of α₅β₁ (53). Several research groups have investigated the partitioning of raft-associated proteins using giant unilamellar vesicles and giant plasma membrane vesicles. Results show that the relative preference for ordered phases is higher for these proteins than for nonraft-associated proteins, and this preference increases upon addition of cross-linking agents (25–27,54). Our system differs from those approaches in that it is planar and investigates the oligomerization state along with sequestration in the absence of artificial cross-linking.

Determination of degree of oligomerization

Our previous data showed that ligand addition causes substantial changes in E_raft, which raises the possibility that these changes are accompanied by similar substantial changes in integrin oligomerization state. Ligands are known to induce clustering in membrane proteins other than integrins (55–57). Moreover, clustering (through agents such as GM1 or a cross-linking antibody) is known to induce raftophilicity in integrins (20,58,59). To investigate whether ligand addition by itself induces a change in oligomerization state separately from any change that might be induced by a change in lipid phase, we constructed a series of type II bilayers, which do not phase-separate, with 0, 5, and 30 mol % CHOL. We probed the oligomerization state of α_vβ₃ and α₅β₁ by analyzing the PCH data acquired in these six systems before and after ligand binding. The PCH curves and best fits for species brightness are provided in Fig. S3 and Fig. S4. They show that the integrins were primarily monomers both before and after ligand binding, and the best fits for brightness for these systems in each case were within 15% of the best fit for the brightness of the MAbs in solution (Fig. S4). The results for oligomerization state as deduced from solutions to the PCH algorithm are shown in Fig. 4. For the α_vβ₃ system, X_dimer is below 5% for the CHOL free bilayer, but increases to 5–10% for 5 mol % CHOL and then to ∼12% for the 30 mol % CHOL bilayer. The α₅β₁ system shows the same trend, with possibly higher X_dimer (2–7%) found for 0% CHOL. This finding is interesting in light of the observation that reduction of CHOL levels leads to a reduction in integrin functioning as observed by reduced cellular adhesion capabilities (13). Of more importance, there is no statistical difference in oligomerization state before (light bars) and after (dark bars) ligand binding, with the possible exception of a moderate increase in oligomerization after ligand binding in the 5 mol % CHOL α_vβ₃ system.

Figure 4.

Fraction of dimers found through PCH analysis for α_vβ₃ (A) and α₅β₁ (B) before (light bars) and after (dark bars) ligand binding in type II bilayers. These data show that increasing CHOL increases oligomerization for both α_vβ₃ and α₅β₁, but only to moderate levels, and ligand binding has no statistically measurable effect on dimerization.

After characterizing integrin oligomerization in type II bilayers, we conducted corresponding experiments on raft-mimicking type I bilayers, utilizing the dual EPI/FFS setup to acquire FFS data in l_o and l_d phases for both α_vβ₃ and α₅β₁ before and after ligand addition for the purpose of analyzing oligomerization state in this system. The results of these studies are shown in Fig. 5, which include the PCH data sets as well as the findings for brightness relative to MAbs in solution and X_dimer. PCH analysis shows that the primary brightness of the integrins relative to the fluorescent MAbs in solution was 77 ± 9% before ligand binding and 81 ± 9% after ligand binding, and thus was essentially the same before and after ligand binding (Fig. 5 E). The rate of dimerization was moderate, between 5 and 20 mol % for both α_vβ₃ and α₅β₁ (Fig. 5 F), thus mirroring the results found in the type II bilayers. Using our system and the PCH algorithm, we were able to sensitively identify the oligomerization state, and found that raft sequestration due to monomeric, nonclustered ligands does not induce oligomerization in either α_vβ₃ or α₅β₁ integrins.

Figure 5.

(A–D) PCH curves for α_vβ₃ (A and C) and α₅β₁ (B and D) before (light markers) and after (dark markers) ligand binding in both l_o phase (A and B) and l_d phase (C and D), along with PCH curves for MAbs for integrins in solution (open markers). MAbs data were acquired twice: at the time of initial PCH acquisition (before ligand binding) and at the time of subsequent PCH addition (after ligand binding). Dotted lines are best-fit curves from the PCH algorithm. (E) Brightness compared with MAbs in solution. (F) Fraction of dimers found through PCH analysis of α_vβ₃ (left) and α₅β₁ (right) before (light bars) and after (dark bars) ligand binding in l_d and l_o phases.

Interestingly, our findings are in good agreement with an early study involving octyl-glucoside, in which Hantgan et al. (60) quantified the average molecular weight of the integrin α_IIbβ₃ through centrifugation before and after addition of an RGD peptide ligand-mimetic. Their study showed only a 10% increase in molecular weight after ligand addition, and thus did not detect significant integrin oligomerization upon ligand addition either. Our PCH data are also supported by an elegant cell study that revealed that ligand addition to α_vβ₃ or other forms of α_vβ₃ activation was not capable of inducing clustering in the absence of cytosolic-integrin linkages (58). These findings, as well as our data, imply that α_vβ₃ integrin-ligand binding alone is insufficient for integrin clustering or oligomerization. Investigators have studied α₅β₁ clustering by monitoring the difference in strength of cellular adhesion (a proxy for integrin clustering) of magnetic beads coated with different concentrations of FN (21) or polymers linked with 1.7, 3.6, or 5.4 RGD peptides (20). In both cases, cellular adhesion was significantly stronger (per ligand attached) for clustered ligands than for monovalent ligands, indicating that single ligands were not spontaneously forming clusters.

The combined integrin sequestering and PCH data clearly indicate that ligand binding does not affect X_dimer for either α_vβ₃ or α₅β₁ in phase-separating type I lipid mixtures. In other words, the observed ligand-mediated changes of α_vβ₃ and α₅β₁ sequestering reported in Figs. 1–3 are not caused by changes in receptor oligomerization state. Moreover, our study demonstrates that the change in the fraction of proteins that dimerize before and after ligand binding (≤5% difference in X_dimer on ligand addition) is far less than the fraction of proteins that migrated from disordered to ordered lipid phases for both the α₅β₁ system, where 27% migrated to induce an even distribution, and the α_vβ₃ system, where 54% migrated to the l_o phase to induce a clear preference for the l_o phase. This platform therefore gives us the ability to sensitively distinguish two separate aspects (i.e., raft-association and oligomerization state) and to conclude that although ligand binding is sufficient to induce raft association, it is not directly implicated in oligomerization.

Our data suggest that the observed protein sequestering is due to ligand-induced conformational changes of integrins influencing integrin-lipid interactions. It is well known that ligand addition causes substantial structural changes to both the ectodomains and transmembrane domains of integrins. Electron microscopy studies have analyzed the headgroup conformation of the EC domains of both α_vβ₃ and α₅β₁ in solution before and after quantitative exposure to RGD peptides designed to mimic ECM ligands (61,62). Both α_vβ₃ and α₅β₁ integrin fragments started in the resting, bent conformations before they were exposed to ligands. After ligand exposure, 98% of the α_vβ₃ headgroups adopted an open conformation, even in the absence of the known activating factor, Mn²⁺ (61), and 25% of the α₅β₁ protein extracellular domains switched to the open conformation (62). The position of the headgroups in relation to the membrane also changes: integrins in the bent, resting position have the RGD binding pocket near the plasma membrane, but when activated, the RGD binding pocket is thought to straighten out and pull far away from the membrane (63). These changes in the ectodomain may influence the observed change in sequestration directly by altering the ectodomain-membrane interface, or indirectly by inducing changes in the integrin transmembrane domains. Furthermore, studies on mutations of integrins indicate that the transmembrane domains of the α and β subunits are associated with substantial conformational changes in response to external stimuli during processes associated with integrin outside-in and inside-out signaling (64). Such conformational changes likely expose different residues and increase the number of residues in the β subunit that reside within the lipid membrane, thus changing the tilt angle of the β subunit (65). This process may affect the mismatch between the integrin transmembrane domain and the hydrophobic region of the lipid bilayer (66).

Conclusion

This study demonstrates that polymer-tethered planar model membranes are sensitive tools for studying transmembrane proteins in different lipid environments and acquiring quantitative data regarding the partitioning preference and oligomerization state of these systems. In this study, we investigated integrin preferences for raft-like and nonraft-like lipid environments using a bilayer phase-separated into l_o and l_d phases. We found that two members of the integrin family, α_vβ₃ and α₅β_1, partition preferentially into the l_d phase in their native state. However, addition of ECM ligands induced a quantifiable change in the partition preferences: FN caused α₅β₁ to lose preference and partition equally between ordered and disordered states, and VN caused α_vβ₃ to partition as strongly in the l_o phase as it did in the l_d phase. In addition, using PCH, we were able to quantify the degree of oligomerization of the proteins, and found that most of the proteins, for both systems, were monomers, and addition of ligands did not change the oligomerization state. We further investigated the effect of CHOL on the oligomerization state of the proteins, and found for both systems that increasing CHOL levels increased the oligomerization state, but only moderately, and again, addition of ECM ligands had no effect on the oligomerization state. Our data show that the experimental platform presented here enables the isolation of different factors that describe the effect of lipid rafts in regulating membrane proteins without the use of confounding cross-linking agents. This methodology can be expanded to study the incorporation of other membrane proteins and cofactors to the model system, as well as to study the effect of asymmetric bilayers with lipid rafts contained in only the inner or outer leaflet. Recent work showing that α₅β₁ appears to recruit CHOL to the outer leaflet in plasma membranes (67) makes this a fascinating avenue for investigation. The methodology can further be combined with other approaches involving the use of polymer-tethered lipid bilayers, such as that used by Purrucker et al. (30) to investigate the cell adhesion strength of integrins incorporated in polymer-tethered model systems. Finally, we hope that this work will further membrane protein molecular-dynamics simulations incorporating CHOL.