Mode of Ezrin-Membrane Interaction as a Function of PIP₂ Binding and Pseudophosphorylation

Abstract

Ezrin, a protein of the ezrin, radixin, moesin (ERM) family, provides a regulated linkage between the plasma membrane and the cytoskeleton. The hallmark of this linkage is the activation of ezrin by phosphatidylinositol-4,5-bisphosphate (PIP₂) binding and a threonine phosphorylation at position 567. To analyze the influence of these activating factors on the organization of ezrin on lipid membranes and the proposed concomitant oligomer-monomer transition, we made use of supported lipid bilayers in conjunction with atomic force microscopy and fluorescence microscopy. Bilayers doped with either PIP₂ as the natural receptor lipid of ezrin or a Ni-nitrilotriacetic acid-equipped lipid to bind the proteins via their His₆-tags to the lipid membrane were used to bind two different ezrin variants: ezrin wild-type and ezrin T567D mimicking the phosphorylated state. Using a combination of reflectometric interference spectroscopy, atomic force microscopy, and Förster resonance energy transfer experiments, we show that only the ezrin T567D mutant, upon binding to PIP₂-containing bilayers, undergoes a remarkable conformational change, which we attribute to an opening of the conformation resulting in monomeric protein on the lipid bilayer.

Introduction

The proteins of the ezrin, radixin, moesin (ERM) family are important regulators of plasma membrane-cytoskeleton connections that function in many cellular processes, ranging from cell shape regulation to membrane traffic and cell migration (1, 2, 3, 4). ERM proteins share a common structure composed of three distinct domains: 1) the N-terminal membrane-binding domain, which has similarities to a domain in protein 4.1 and is termed the FERM domain (1); 2) the central α-helical domain; and 3) the C-terminal actin-binding domain. One member of this family, ezrin, is particularly enriched in intestinal microvilli and in filopodia (5, 6). Like all ERM proteins, it can exist in at least two different states, closed (inactive) or open (active) (7). In its inactive dormant state, ezrin adopts an autoinhibited conformation due to intramolecular interactions of its N-terminal FERM domain (N-terminal ERM association domain (N-ERMAD)) with its C-terminal ERM association domain (C-ERMAD) (8). Inactive ezrin proteins in their closed conformation have been shown to form oligomers through N-ERMAD-C-ERMAD intermolecular interactions. These oligomers have been localized in the cytoplasm and at the membrane (9, 10, 11). Activation of ezrin has been proposed to follow a two-step mechanism. The first activation occurs via phosphatidylinositol-4,5-bisphosphate (PIP₂) binding at the membrane (12, 13). Through this PIP₂ binding, threonine 567 in C-ERMAD becomes accessible for phosphorylation by Rho-kinase and several PKC isoforms (1, 14, 15). This phosphorylation of threonine 567 represents the second activation step and induces a transition from inactive ezrin oligomers to active membrane-associated monomers that act as cross-linkers between the plasma membrane and the actin cortex (10). Although the PIP₂-binding site is located in N-ERMAD and is accessible in the closed conformation (12), F-actin binding occurs through C-ERMAD and requires full ezrin activation (16). Likewise, binding sites for several integral membrane proteins that function as ezrin acceptors and are located in N-ERMAD are only accessible in activated ezrin (1).

The interactions between ezrin and membranes, and between F-actin and ezrin, are difficult to elucidate in a cellular environment because the presence of hundreds of actin- and ERM-binding proteins may hide the true contribution of ezrin. This complexity can be reduced by employing biomimetic systems comprising only a limited number of components. Such systems have been introduced into the field of ERM proteins in an attempt to understand their molecular behavior in complex systems (17, 18, 19, 20, 21). In addition, in vitro experiments using well-defined lipid membranes and purified proteins can provide quantitative information about protein-membrane interactions and conformational changes associated with the activation process of ezrin. Three main types of biomimetic membranes are used to study ERM-membrane interactions: large unilamellar vesicles (12, 22, 23, 24), giant unilamellar vesicles (25), and supported lipid bilayers (SLBs). In particular, membranes attached to a solid substrate have been shown to be well suited for monitoring the details of ezrin binding (26, 27), as well as the interaction of ezrin with actin filaments (28).

In this study, we made use of SLBs to investigate the organization of ezrin on lipid membranes as a function of its activation state and thereby elucidate details of the activation process and the proposed concomitant oligomer-monomer transition. We employed two different ezrin variants (ezrin wild-type (wt) and ezrin T567D mimicking the phosphorylated protein (10, 16, 29)) in combination with SLBs doped with either PIP₂ as the natural receptor lipid of ezrin or a Ni-nitrilotriacetic acid (Ni-NTA)-equipped lipid to bind the proteins via their His₆-tags to the membrane. By means of reflectometric interference spectroscopy (RIfS), atomic force microscopy (AFM), and Förster resonance energy transfer (FRET) experiments, we found that ezrin T567D bound via PIP₂ to the SLBs experienced a marked conformational change, most likely an opening of the conformation, which resulted in monomeric protein on the membrane.

Materials and Methods

Materials

L-α-phosphatidylinositol-4,5-bisphosphate (PIP₂, purified from porcine brain with a fatty acid composition primarily composed of 18:0, 18:1, and 20:4 acyl chains), 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl] Ni salt (DOGS-Ni-NTA), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), and 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) were obtained from Avanti Polar Lipids (Alabaster, AL). Silicon wafers were purchased from Silicon Materials (Kaufering, Germany).

Protein purification

N-ERMAD, ezrin wt, and ezrin T567D were obtained by recombinant expression in Escherichia coli (strain BL21(DE3)pLysS) as described previously (27). Briefly, E. coli cells transformed with bacterial expression vectors pET28a+ (Novagen, Madison, WI) encoding N-ERMAD or ezrin derivatives with an N-terminal His₆-tag were grown to an OD₆₀₀ of 0.6. Protein expression was induced by adding 1 mM isopropyl-β-D-thiogalactopyranoside. After 3 h, the cells were harvested by centrifugation (4000 × g, 20 min, 4°C) and the pellet was resuspended in lysis buffer (40 mM HEPES, 20 mM imidazole, 300 mM NaCl, 1 mM EDTA, 10 mM β-mercaptoethanol, protease inhibitor cocktail (Complete; Roche Diagnostics, Basel, Switzerland), pH 7.4). To complete lysis, the suspension was sonicated on ice. The bacterial lysate was clarified by centrifugation (100,000 × g, 1 h, 4°C) and the supernatant was applied to a Ni-nitrilotriacetic acid (Ni-NTA) agarose column (Qiagen, Hilden, Germany) equilibrated with lysis buffer. The column was washed twice with lysis buffer supplemented with 30 mM imidazole, pH 7.4, and 50 mM imidazole, pH 7.4, respectively. Protein was eluted using buffer A (20 mM Tris/HCl, 50 mM KCl, 0.1 mM EDTA, 0.1 mM NaN₃, pH 7.4) containing 250 mM imidazole, dialyzed against imidazole-free buffer A, and stored at 4°C until use. Protein concentration was determined by UV/Vis spectroscopy using an extinction coefficient of ε₂₈₀ = 66,900 M⁻¹ cm⁻¹ for ezrin wt and T567D, and ε₂₈₀ = 52,400 M⁻¹ cm⁻¹ for N-ERMAD.

Protein labeling

C-terminal enhanced green fluorescent protein (eGFP)-tagged ezrin derivatives containing an N-terminal His₆-tag (ezrin-eGFP and ezrin T567D-eGFP) were expressed in E. coli (strain BL21(DE3)pLysS) and purified via the His₆-tag on a Ni-NTA column as described above, with the following modifications: To reduce proteolysis of the eGFP-tagged constructs, bacterial cell lysis was performed using a French cell press (SLM Aminco). Moreover, 500 mM imidazole, 150 mM NaCl, 10 mM β-mercaptoethanol, and 1 mM phenylmethylsulfonyl fluoride, pH 7.4, were used to elute the proteins from the column. Purified ezrin-eGFP and ezrin T567D-eGFP were chemically labeled at the N-terminal amino group using Cy3-mono-reactive N-hydroxysuccinimide (NHS)-ester (GE Healthcare, Pittsburgh, PA). Proteins were first dialyzed against labeling buffer (10 mM PIPES, 150 mM NaCl, pH 6.5) and then mixed with Cy3-NHS at a molar ratio of 1:1 and incubated for 3 min at room temperature in the dark. The slightly acidic pH favors the reaction of Cy3-NHS with the N-terminal amino group as compared with the side-chain amino groups. Excess free dye was removed by size-exclusion chromatography (Sephadex columns Illustra NAP 10; GE Healthcare). The dye/protein ratio of the product was determined using UV/Vis spectroscopy, taking the absorption value at 280 nm (protein) and 550 nm (Cy3) into account. A dye/protein ratio in the range of 0.6–1 was determined for all preparations used in this study.

Vesicle preparation

Stock solutions of the respective lipids were prepared in chloroform at concentrations ranging from 0.5–10 mg/mL except for PIP₂, which was dissolved in a mixture of chloroform/methanol/water (20:9:1) at 1 mg/mL. Lipid stock solutions (0.4–0.6 mg of total lipid) were mixed in a test tube preloaded with 200 μL chloroform at the desired molar ratio. Fluorophores were added as indicated. The organic solvent was evaporated with a gentle stream of nitrogen at a temperature above the lipid gel-fluid phase transition. To remove residual solvent, the lipid film was further dried under vacuum for 3 h at the same temperature. Lipid films were stored at 4°C until use. A lipid film was rehydrated by adding 0.5–1.0 mL of buffer solution and incubating for 20 min. Multilamellar vesicles were then prepared by vortexing for 3 × 30 s at 5 min intervals. The suspension of multilamellar vesicles was transferred to an Eppendorf cup and sonicated for 2 × 15 min using an ultrasonic homogenizer (Sonopuls HD2070, resonator cup; Bandelin, Berlin, Germany) to obtain small unilamellar vesicles (SUVs).

Preparation of SLBs

Silicon substrates were rinsed thoroughly with isopropanol and water. An aqueous solution of NH₃ and H₂O₂ (H₂O/NH₃/H₂O₂, 5:1:1, 20 min at 70°C) was used to remove organic contaminations and render the surface of the silicon substrate hydrophilic. Additional hydrophilization was achieved by oxygen plasma treatment for 2 min. For RIfS analysis, silicon substrates with a 5-μm-thick SiO₂ layer were used, and for AFM and fluorescence experiments, silicon substrates with a 100-nm-thick SiO₂ layer were employed. To form SLBs, the hydrophilized substrate was directly treated with a freshly prepared SUV suspension (0.2 mg/mL) in the corresponding buffer solution depending on the receptor lipid (27). After incubation for 2 h at room temperature or at 55°C in the case of DPPC-containing membranes, the SLB obtained was rinsed with buffer to remove the remaining lipid material from the solution.

Fluorescence microscopy and FRET analysis

Fluorescence imaging and FRET analysis were performed with an LSM 710 confocal fluorescence laser scanning microscope (Carl Zeiss, Jena, Germany). The microscope was equipped with a water immersion 63× objective with a numerical aperture of NA = 1 (W Plan Apochromat; Carl Zeiss) and operated with the software Zen 2010. An Ar laser (488 nm) and a diode-pumped solid-state laser (561 nm) were used for excitation of eGFP and Cy3, respectively. Fluorescence emission was recorded in the wavelength range of 493–550 nm for eGFP and 575–681 nm for Cy3.

For FRET analysis, 1.5 μM Cy3-ezrin-eGFP or 1.5 μM Cy3-ezrinT567D-eGFP was bound to SLBs composed of POPC/PIP₂ (96:4) in buffer A or POPC/DOGS-Ni-NTA (96:4) in buffer A lacking EDTA for 12–24 h at 4°C. This protein concentration ensures that full protein surface coverage is reached (28). To be able to observe intermolecular FRET only, we added single-labeled ezrin constructs (Cy3-ezrin + ezrin-eGFP or Cy3-ezrin T567D + ezrin T567D-eGFP) mixed in a 1:1 ratio (0.75 μM each) for 12–14 h at 4°C. Under the assumption that all proteins bind with roughly the same affinity and the same kinetics, this procedure should result in the same ratio of donor and acceptor (R_DA = 1) on the membrane surface (30). After incubation, the supported protein-decorated lipid bilayers were rinsed with the corresponding buffer. FRET between the C-terminal donor eGFP and the N-terminal acceptor Cy3 was recorded by acceptor photobleaching using laser irradiation at 561 nm of a defined region of interest (ROI) on the membrane surface. The intensity of the donor fluorescence before (I_{D_prebleach}) and after (I_D__postbleach) acceptor bleaching in the ROI was monitored to quantify the FRET efficiency with the FRETcalc plugin software of ImageJ. The FRET efficiency is defined as

$$\text{\%FRET}=\frac{I_{\text{D\_postbleach}}-I_{\text{D\_prebleach}}}{I_{\text{D\_postbleach}}}100,$$ (1)

showing that the FRET efficiency is normalized to the surface density of the donor fluorophores.

RIfS

RIfS was employed to monitor the formation of SLBs and subsequent protein binding in a time-resolved manner (27). The experimental setup used in this work has been described elsewhere (31). Briefly, RIfS experiments were performed using a NanoCalc-2000-Vis/NIR spectrometer (Ocean Optics, Dunedin, FL). Spectra were recorded every 2 s and data were evaluated using a MATLAB (The MathWorks, Natick, MA) tool.

AFM

AFM images were taken using NanoWizard II and III systems (JPK Instruments, Berlin, Germany) and an MFP-3D instrument (Asylum Research, Santa Barbara, CA) with CSC37 (k = 0.65 mN/m, f = 41 kHz; MikroMasch, Tallinn, Estonia) or MSNL-10 (k = 0.6 mN/m, f = 125 kHz; Bruker, Camarillo, CA) cantilevers, respectively, at room temperature in intermittent contact mode (scan speed: 10 μm s⁻¹, 512 × 512 pixels). The AFM images were analyzed using the programs Gwyddion and MATLAB. For the height analysis, 16-bit grayscale images were exported and baseline corrected. The membrane interface was detected in each image and set to 0 nm. A MATLAB script making use of a peak-finding algorithm was employed to determine the height differences between the elevated protein structure and the lipid bilayer in the topography image line by line. The detected height differences per image were cast into a histogram and the average height of the membrane-bound protein was determined as the maximum of the histogram. A histogram obtained from all analyzed AFM images provides the average protein height for a particular binding scenario.

Results

Analysis of ezrin-lipid interactions by RIfS

RIfS is an optical technique that allows label-free investigations of protein binding to a specific receptor lipid embedded in SLBs (27, 31, 32, 33, 34, 35). The adsorption process is detected as an increase in optical thickness, OT = n d, i.e., the product of the refractive index n and the physical layer thickness d. RIfS experiments were employed to analyze whether ezrin wt or ezrin T567D adsorbed to either the receptor lipid DOGS-Ni-NTA or PIP₂ results in a difference in OT. Detection of a putative difference in the change in OT might be an indication of a conformational change of the membrane-bound protein. SLBs were prepared on silicon/silicon dioxide substrates by spreading of SUVs. Membranes were composed of POPC/PIP₂ for binding of ezrin wt or ezrin T567D to the natural receptor PIP₂, or composed of DOPC/DOGS-Ni-NTA to bind the proteins via the N-terminal His₆-tag. By using these immobilization strategies in combination with the two different ezrin variants, we aimed to address the individual contributions of the activating factors PIP₂ and phosphorylation, respectively. We used receptor lipid contents of 2 mol % and 4 mol %, reflecting a reasonable physiological range of PIP₂ in the plasma membrane.

The spreading of DOPC/DOGS-Ni-NTA SUVs to form SLBs (Fig. S1 in the Supporting Material) results in a change in OT of (6.3 ± 0.2) nm for DOPC/DOGS-Ni-NTA (98:2) bilayers (n = 17) and (6.2 ± 0.6) nm for DOPC/DOGS-Ni-NTA (96:4) bilayers (n = 9). Assuming n_lipid = 1.5 (36), a physical membrane thickness of d_mem = 4.2 nm is calculated with the relation ΔOT = n_lipid d_mem, which is in good agreement with literature values reporting a thickness of 4.1 nm for a DOPC bilayer at room temperature (37). Similarly, the spreading of POPC/PIP₂ SUVs (Fig. S1) results in an OT change of (6.4 ± 0.5) nm (2 mol % PIP₂, n = 22) and (6.3 ± 0.6) nm (4 mol % PIP₂, n = 11), respectively. These ΔOT values translate to a bilayer thickness of 4.1 nm, in agreement with the bilayer thickness of (3.98 ± 0.08) nm previously determined for a POPC bilayer by small-angle neutron and x-ray scattering (38).

Ezrin wt or ezrin T567D was added to these planar bilayers, and the change in OT was again monitored (Fig. 1). The two proteins produced a significant increase in OT on both DOGS-Ni-NTA-doped membranes and PIP₂-containing ones. In the absence of receptor lipids, no ezrin binding was observed (not shown). Upon rinsing with buffer, only a small amount of protein desorbed from the surface, indicating that the majority of protein remained irreversibly bound to the membrane. The phenomenon of irreversibly bound ezrin to receptor lipid-doped SLBs was investigated in more detail in previous experiments (26, 27). The experiments showed that not more than 20% of bound ezrin desorbed from the membrane as a result of lateral protein-protein interactions leading to protein clusters on the membrane surface, as visualized by AFM. Based on these experiments and theoretical considerations, it was concluded that individual ezrin molecules are reversibly bound to PIP₂; however, once they are bound, subsequent lateral protein-protein interactions lead to protein clusters on the membrane, preventing desorption of the single protein. The monitored ΔOT values as a function of receptor lipid and receptor density are given in Fig. 2.

Figure 1.

Binding of ezrin wt and ezrin T567D to SLBs composed of DOPC/DOGS-Ni-NTA or POPC/PIP₂ (96:4), analyzed by RIfS. (a) Addition of 0.8 μM protein solution leads to an increase in OT. (b) After saturation, the surface is rinsed with buffer.

Figure 2.

Results of the RIfS experiments. The change in OT is plotted for ezrin wt (gray) and ezrin T567D (red) bound to either 2 or 4 mol % DOGS-Ni-NTA (crosswise striated) or PIP₂ (longitudinally striated). Three to eight independent experiments were performed in each case. The error bars are the standard deviation of the mean (^∗p < 0.04, ^∗∗p < 0.01; ns, not significant; two-sample (nonpaired) Student’s t-test). To see this figure in color, go online.

The observed changes in OT are a result of the actual protein height and the surface coverage, and it is obvious that an increase in receptor density increases the protein surface coverage and hence the ΔOT. On DOGS-Ni-NTA, the ΔOT values for ezrin wt and ezrin T567D are almost identical, whereas on PIP₂ the ΔOT is reduced by a factor of 5 for the ezrin T567D mutant but only by a factor of 2 for ezrin wt. This might be the result of a different protein height on the lipid bilayer surface and/or a different lipid bilayer surface coverage owing to a conformational change of the protein upon binding to PIP₂. To elucidate this in more detail, we performed AFM studies that enabled us to determine the protein height on SLBs precisely without interference from the surface coverage.

Structural organization of membrane-bound ezrin analyzed by AFM

Planar membranes composed of DPPC and doped with either DOGS-Ni-NTA or PIP₂ were prepared on silicon/silicon dioxide substrates by spreading of SUVs. DPPC was chosen because the resulting planar lipid bilayers would be in the gel phase at room temperature and thus less prone to undergo force-dependent changes in their structure.

The success of the membrane preparations was monitored by fluorescence microscopy (Fig. S2) and AFM as depicted in Fig. 3 A. A planar lipid bilayer with irregular defects was monitored. The height difference between the lipid bilayer and the defects was ∼5–6 nm, which is characteristic for a DPPC lipid bilayer (Fig. 3 A, line profile) (37). The same result was observed for DPPC bilayers doped with DOGS-Ni-NTA (not shown). Although the crystal structure of full-length ezrin has not yet been solved, the crystal structure of the N-terminal FERM domain (N-ERMAD) of ezrin is known (39). Hence, to validate our approach for determining protein height by AFM, we first conducted experiments with the isolated N-ERMAD of ezrin. A DPPC/PIP₂ (97:3) planar lipid bilayer was incubated overnight with 0.5 μM N-ERMAD and then analyzed by AFM. A netlike protein structure was found on the lipid bilayer (Fig. 3 B1). The protein did not significantly bind to the membrane defects, which were still clearly visible. The protein height on the lipid bilayer was then determined as described in the Materials and Methods section, and a height histogram was obtained.

Figure 3.

Atomic force micrographs of an SLB before and after protein incubation. (A) Atomic force micrograph of a lipid bilayer (DPPC/PIP₂, 97:3) on a silicon/silicon dioxide substrate (A1). The corresponding line profile (A2) along the blue line shown in the topography image (A1) demonstrates the relative height of a single bilayer (interface set to 0 nm) with respect to a membrane defect. (B) Atomic force micrograph of a lipid bilayer (DPPC/PIP₂, 97:3) that was incubated with N-ERMAD (0.5 μM, 15 h) and subsequently rinsed with buffer (B1). The height profile of the protein clusters (B2) along the blue line shown in the topography image (B1) demonstrates the height difference between the lipid bilayer (interface set to 0 nm) and the adsorbed protein clusters. Scale bars, 1 μm. To see this figure in color, go online.

The mean value of the height of N-ERMAD bound to a PIP₂-doped DPPC lipid bilayer was determined to be t_{N-ERMAD (PIP2)} = (3.3 ± 0.9) nm. The same height of N-ERMAD was obtained on a DOGS-Ni-NTA-doped DPPC membrane (t_{N-ERMAD (Ni-NTA)} = (3.4 ± 1.3) nm). These values are in good agreement with the heights expected from the crystal structure of N-ERMAD (see Fig. 9) (39), thus validating our approach for height determination.

Figure 9.

Crystallographic dimensions of N-ERMAD composed of three subdomains (F1, F2, and F3). The top image shows the minimum and maximum dimensions of N-ERMAD. The bottom image shows the possible orientation of N-ERMAD bound to PIP₂, resulting in the observed height of 3.4 nm. To see this figure in color, go online.

We next analyzed the structural organization and height of ezrin wt and the pseudophosphorylated variant ezrin T567D, either bound to PIP₂ via the PIP₂-binding site in the N-terminal FERM domain or bound to DOGS-Ni-NTA via its N-terminal His₆-tag. In all cases, a netlike structure of proteins on the membrane surfaces was observed, with an average surface coverage of 40–70% (Figs. 4 and S3). From the AFM images, the height distribution of the bound proteins was determined. For ezrin wt bound to a DOGS-Ni-NTA-containing lipid bilayer, an average height of t_{ezrin wt (Ni-NTA)} = (3.5 ± 0.8) nm was calculated. The average height of ezrin T567D on a DOGS-Ni-NTA-doped membrane was t_{ezrin T567D (Ni-NTA)} = (3.2 ± 0.9) nm, which is very similar to the height obtained for the wt protein.

Figure 4.

Atomic force micrographs of SLBs (1) composed of DPPC/PIP₂ (97:3) on silicon/silicon dioxide substrates, which were incubated with (A) ezrin wt (0.5 μM) or (B) ezrin T567D (0.5 μM). The SLBs were rinsed with buffer before imaging. The corresponding line profiles of protein clusters (2) marked in the topography image (blue line) show a significantly lower protein height for ezrin T567D. Scale bars: 1 μm. To see this figure in color, go online.

These values are almost the same as those obtained for N-ERMAD and indicate that the C-ERMAD and the central α-helical domain of ezrin contribute only to a minor extent to the overall height of the protein on the lipid bilayer. The same analysis was performed for ezrin wt and the T567D mutant bound to PIP₂-doped membranes. Again, for ezrin wt bound to PIP₂ (Fig. 5 A), an average height of t_{ezrin wt (PIP2)} = (3.0 ± 0.4) nm was determined. However, in the case of ezrin T567D bound to its natural receptor lipid, PIP₂, the average height was strikingly different (Fig. 5 B). A height distribution of t_{ezrin T567D (PIP2)} = (1.6 ± 0.3) nm was determined, which was lower by a factor of 2 than the protein heights obtained in the other scenarios (Fig. 5 C). Therefore, we conclude that the mode of ezrin T567D binding to PIP₂-doped SLBs was significantly altered compared with all of the other investigated combinations and might be the result of a fully open conformation, which is proposed to result in monomeric ezrin molecules.

Figure 5.

Summary of results obtained after height analysis of atomic force micrographs. (A and B) Histograms obtained for (A) ezrin wt bound to DPPC/PIP₂ (97:3) and (B) ezrin T567D bound to DPPC/PIP₂ (97:3). (C) Summary of protein heights determined by AFM images according to the procedure described in the Materials and Methods section for ezrin wt (gray), ezrin T567D (red), and N-ERMAD (green) bound to PIP₂ (longitudinally striated) or DOGS-Ni-NTA (crosswise striated). The error bars are the standard deviation of the mean. The significance of the data was analyzed with Student’s t-tests (^∗∗∗p < 0.001; ns, not significant). To see this figure in color, go online.

To further investigate this, we performed FRET experiments to analyze the interactions between the N- and C-terminal ends of the proteins, and obtain information about the intramolecular (closed or open conformation) and intermolecular (monomeric or oligomeric) self-association of ezrin (40).

Analysis of a conformational change in membrane-bound ezrin using FRET

In a previous study, Zhu et al. (11) showed that FRET can be used to visualize interactions between the N- and C-terminal ends of proteins. In that study, they employed an ezrin construct that contained an N-terminal yellow fluorescent protein and a C-terminal cyan fluorescent protein serving as the FRET donor/acceptor pair. Here, we generated ezrin constructs with only one large GFP moiety (eGFP) at the C-terminus and labeled the N-terminus with an organic dye, Cy3 (Cy3-ezrin-eGFP), which is known to form a donor-acceptor pair with eGFP. The fluorescence spectra of the protein constructs are shown in Fig. 6. Both the wt and the T567D mutant show a FRET peak in solution, indicative of an intramolecular FRET due to the close proximity of the N- and C-terminal ends. Of note, the same spectra were obtained for the wt and mutant proteins, suggesting that even the T567D mutant resides in a closed conformation in which the N- and C-termini are close enough and in the correct orientation to produce FRET. To further assess intramolecular FRET, we treated the protein constructs with the protease calpain I, which is known to cleave ezrin, while leaving the GFP intact (11, 41). The treatment resulted in a diminished FRET peak, indicating a partial separation of the eGFP and Cy3 dye (Fig. S4).

Figure 6.

(A and B) Fluorescence spectra of (A) ezrin wt and (B) ezrinT567D. Ezrin-eGFP was excited at 488 nm (green line), and Cy3-ezrin-eGFP was excited at 488 nm (blue line) and 550 nm (red line, fluorescence intensity divided by 5). The black line is the difference spectrum of the spectra shown in blue and green. To see this figure in color, go online.

Next, the labeled proteins were bound to SLBs containing either DOGS-Ni-NTA or PIP₂ as the receptor lipid. All proteins bound uniformly to the lipid bilayers, as visualized by fluorescence microscopy of the respective specimen (one example for Cy3-ezrin wt-eGFP is shown in Fig. 7). We then recorded the FRET in a defined ROI using acceptor photobleaching and calculated the respective FRET efficiency (Eq. 1) for each case using ImageJ (see Materials and Methods section). We used 1:1 mixtures of Cy3-ezrin and ezrin-eGFP as controls so that we could solely monitor intermolecular FRET occurring under the conditions chosen. The box plot in Fig. 8 summarizes the results. The single-labeled 1:1 mixtures of proteins show a significantly lower FRET efficiency than the double-labeled constructs. Assuming the same (full) protein surface coverage for single- and double-labeled protein constructs as a result of the chosen high protein concentration, the FRET efficiency would be the same if only intermolecular FRET occurred, even though the donor and acceptor density would be doubled in the case of double-labeled proteins (see Eq. 1). However, since we see a significantly larger FRET efficiency in the case of double-labeled proteins, we conclude that the donor and acceptor are closer to each other, which might indicate that we also observe intramolecular FRET.

Figure 7.

Acceptor bleaching of Cy3-ezrin wt-eGFP bound to a SLB composed of DOPC/DOGS-Ni-NTA (96:4). (A) eGFP, fluorescence, (B) Cy3 fluorescence. The yellow circle 1 marks the ROI after bleaching of Cy3, circle 2 marks the reference ROI. Scale bar: 5 μm. (C) Time trace of the bleaching process showing the increase in eGFP fluorescence (green line, 1A) and the decrease in Cy3 fluorescence (red line, 1B), numbers 1 and 2 correspond to bleached ROI and reference ROI, respectively. To see this figure in color, go online.

Figure 8.

Summary of the FRET analysis of SLB-bound ezrin. The double-labeled ezrin constructs (left: wt, Cy3-ezrin wt-eGFP; T567D, Cy3-ezrin T567D-eGFP) and 1:1 mixtures of single-labeled protein constructs (right: wt, Cy3-ezrin wt/ezrin wt-eGFP; T567D, Cy3-ezrin T567D/ezrin T567D-eGFP) were bound to DOGS-Ni-NTA (crosswise striated) or PIP₂ (longitudinally striated) containing lipid bilayers and the FRET efficiency was measured by acceptor bleaching. The solid line in each box is the median and the dot marks the mean value. Each data point is plotted next to the respective box (black/red dots). The FRET efficiency (%FRET, Eq. 1) was calculated using FRETcalc in ImageJ. The statistical significance of the data was analyzed with a nonparametric Mann-Whitney U-test, which showed differences for ezrin T567D with ^∗∗∗p < 0.001, and nonsignificant ones (ns) for ezrin wt. To see this figure in color, go online.

The FRET efficiencies obtained for the ezrin wt double- and single-labeled proteins are very similar and basically independent of the receptor lipid. However, in the case of ezrin T567D, a slightly smaller FRET is observed when the single-labeled proteins are bound to PIP₂. This difference between the FRET efficiencies of the protein bound via DOGS-Ni-NTA or PIP₂ is even more pronounced when the double-labeled protein Cy3-ezrin T567D-eGFP is analyzed. This indicates that the proximity of the N- and C-terminal ends of ezrin T567D is affected upon binding to the lipid PIP₂, which suggests an altered conformation of the protein on PIP₂ as compared with a DOGS-Ni-NTA-containing lipid bilayer.

Discussion

We analyzed the organization and orientation of ezrin wt and the mutant ezrin T567D mimicking the phosphorylated state of the protein on SLBs as a function of its natural receptor lipid PIP₂ by means of RIfS, AFM, and fluorescence microscopy. From our results, we conclude that the binding mode of ezrin T567D to PIP₂ differs considerably from that of the wt protein and also from that of the D-mutant bound to DOGS-Ni-NTA. RIfS results provide a physical thickness of a protein layer that is determined by the height of the layer as well as the surface coverage (27, 31, 32, 42). For ezrin bound to DOGS-Ni-NTA or PIP₂, the change in OT as a result of protein surface coverage is reflected by the increase in ΔOT as a function of increasing receptor lipid content in the membrane. If the refractive index of an adsorbed protein layer is assumed to be n_prot = 1.455 (43), the physical layer thicknesses for the different binding scenarios amount to 0.3–3.3 nm. Obviously, incomplete protein surface coverage impairs the determination of absolute heights by RIfS. However, it is obvious that ezrin T567D bound via PIP₂ shows a remarkably reduced ΔOT compared with the other binding scenarios. This can be explained by either a different surface coverage or a different protein arrangement on the surface leading to an altered protein height.

To analyze this in more detail, we performed AFM experiments that enabled us to determine the protein height without the influence of the surface coverage. Within the error of the AFM experiments, the heights obtained for N-ERMAD, ezrin wt, and ezrin T567D bound to DOGS-Ni-NTA are the same, 3.4 nm on average. The analysis of N-ERMAD served as a benchmark because its crystal structure is known (39), whereas the crystal structure of full-length ezrin has not yet been solved. A height of ∼3.4 nm can indeed be extrapolated from the N-ERMAD structure, as illustrated in Fig. 9 (bottom image).

Almost the same heights were found for N-ERMAD and ezrin wt bound to PIP₂, which indicates that C-ERMAD does not significantly contribute to the overall height of the protein bound to its natural receptor lipid. Interestingly, our analysis revealed the same heights also for ezrin wt and ezrin T567D bound to DOGS-Ni-NTA. Assuming that ezrin T567D would be in an open conformation, one could expect a different binding mode and thus a different height, which was not observed in our AFM experiments. However, Chambers and Bretscher (44) showed that the T567D mutation impairs the association of the N- and C-termini only to a minor extent. They found that even though ezrin T567D carries an additional charge at position 567, it still forms N-ERMAD-C-ERMAD contacts. In solution, both ezrin wt and ezrin T567D form globular monomers and elongated dimers via N-ERMAD/C-ERMAD contacts. Although dimer formation is possible, Chambers and Bretscher (44) also showed that ezrin T567D dimers are less stable than those of the wt protein, and therefore concluded that ezrin T567D spends more time in the open state than the wt protein. Jayasundar et al. (45) corroborated this finding by using contrast-variation small-angle neutron scattering, showing that both ezrin wt and ezrin T567D adopt a closed conformation. Our FRET data obtained in solution support this notion. We obtained the same fluorescence spectra for Cy3-ezrin wt-eGFP and Cy3-ezrin T567D-eGFP, indicating that the two proteins have the same, presumably closed conformation, in which the N- and C-termini are close enough and in the correct orientation to produce FRET.

A different scenario, however, was observed for ezrin T567D bound to PIP₂. In this case, the protein height was significantly reduced to 1.6 nm as compared with the same protein species bound to DOGS-Ni-NTA. Such a height of PIP₂-bound ezrin T567D cannot be explained by the dimensions provided by the crystal structure. Even the shortest distance in the crystal structure would still produce a height of 2.6 nm, which is much larger than the measured 1.6 nm (Fig. 9, top image). An explanation for the observed height of only 1.6 nm is illustrated in Fig. 10. When bound to DOGS-Ni-NTA, ezrin T567D, like ezrin wt, forms a more globular structure that is stabilized by the weakened but still existing N-ERMAD-C-ERMAD interaction. Only upon binding to PIP₂ does the protein open up and adopt an elongated structure, resulting in a less condensed packing on the membrane surface.

Figure 10.

Model of different binding scenarios for ezrin composed of N-ERMAD (F1, F2, and F3) and C-ERMAD on an SLB, demonstrating a more loosely packed ezrin monolayer in the case of the fully activated state (ezrin T567D on PIP₂). To see this figure in color, go online.

Since the atomic force microscope was not capable of resolving individual proteins in this study, we determined an average height value that takes into account the loose packing and the actual height of the protein (3.4 nm). A reduced height was not observed in the case of ezrin wt bound to PIP₂, because some remaining N-ERMAD-C-ERMAD interactions kept the protein in a compact conformation. Thus, both parameters (PIP₂ binding and T567 phosphorylation) are required to significantly weaken the contact between N- and C-ERMAD. To characterize the N-ERMAD-C-ERMAD contact in more detail, we performed FRET experiments. Zhu et al. (11) previously showed that intra- and intermolecular FRET could be monitored for ezrin constructs that contained yellow fluorescent protein at the N-terminus and cyan fluorescent protein at the C-terminus. We observed a significant FRET efficiency for constructs containing Cy3 at the N-terminus and eGFP at the C-terminus (Cy3-ezrin-eGFP) that was larger than the FRET efficiency observed for 1:1 mixtures of single-labeled proteins (Cy3-ezrin and ezrin-eGFP) bound to DOGS-Ni-NTA or PIP₂. As the latter FRET efficiency can only arise from intermolecular FRET, we conclude that in the case of double-labeled proteins, the donor and acceptor come closer to each other, resulting in a larger FRET efficiency, which might be explained by an intramolecular FRET. Interestingly, although the FRET efficiencies were independent of the receptor lipid in the membrane in the case of ezrin wt, some difference was observed for the T567D mutant. In this case, the FRET efficiencies were reduced when the protein was bound to PIP₂ compared with DOGS-Ni-NTA. This reduction was seen for both the 1:1 mixture of single-labeled proteins, where only intermolecular FRET can occur, and the double-labeled proteins, where contributions from intramolecular FRET might be observed. In the case of the 1:1 mixture, this observation can be explained by the notion that the structural rearrangement of ezrin T567D upon PIP₂ binding decreases intermolecular N-ERMAD-C-ERMAD interactions because the proteins are more separated from each other. In the case of the double-labeled construct, the reduction in FRET appears to be even more pronounced. This result can be explained by the different protein organization on the membrane surface in combination with a disruption of the N-ERMAD-C-ERMAD interaction that occurs when both activating events (PIP₂ binding and T567 phosphorylation) take place, resulting in the binding scenario shown in Fig. 10.

Conclusion

An important event in the control of a connection between the cytoskeleton and the plasma membrane is the activation of ezrin, which leads to an increased accessibility of its C-terminal domain harboring the F-actin binding site. Our results demonstrate that monomeric ezrin loosely packed on a planar lipid bilayer is favored if the protein is pseudophosphorylated and bound to its natural receptor, PIP₂. The loosely packed protein configuration on the lipid bilayer allows F-actin to bind readily to the C-terminal domain of ezrin. This is a prerequisite for the protein’s central role in cellular processes (e.g., microvillar generation) and dynamics.

Author Contributions

V.S. and D.G. produced the fluorescently labeled proteins. V.S. and C.K. performed the FRET analysis. C.K., J.B., and J.S. performed RIfS experiments. B.G., J.S., and A.C. performed AFM experiments. I.M. helped with the MATLAB-assisted data analysis. V.G. and C.S. designed the experiments and wrote the manuscript.

Editor: Tobias Baumgart