Dynamic conformational transitions of the EGF receptor in living mammalian cells determined by FRET and fluorescence lifetime imaging microscopy

Authors

  • Iwona Ziomkiewicz,

    1. Laboratory of Cellular Dynamics, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Current affiliation:
    1. Department of Plant and Environmental Sciences, University of Copenhagen, 1871 Frederiksberg, Denmark
    Search for more papers by this author
  • Anastasia Loman,

    1. Department of Neuro- and Sensory Physiology, University Medicine Göttingen, Göttingen, Germany
    Current affiliation:
    1. Centre d'Immunologie, Université Marseille-Luminey, Marseille, France
    Search for more papers by this author
  • Reinhard Klement,

    1. Laboratory of Cellular Dynamics, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    2. Department of Theoretical and Computational Biophysics, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Search for more papers by this author
  • Cornelia Fritsch,

    1. Laboratory of Cellular Dynamics, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Current affiliation:
    1. Dept of Biology, University of Fribourg, 1700 Fribourg, Switzerland
    Search for more papers by this author
    • Iwona Ziomkiewicz and Anastasia Loman contributed equally to the work.

  • Andrey S. Klymchenko,

    1. Laboratoire de Biophotonique et Pharmacologie, UMR 7213 CNRS, Faculté de Pharmacie, Université de Strasbourg, France
    Search for more papers by this author
  • Gertrude Bunt,

    1. Department of Neuro- and Sensory Physiology, University Medicine Göttingen, Göttingen, Germany
    2. DFG Research Center Nanoscale Microscopy and Molecular Physiology of the Brain, Göttingen, Germany
    Search for more papers by this author
  • Thomas M. Jovin,

    1. Laboratory of Cellular Dynamics, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Search for more papers by this author
  • Donna J. Arndt-Jovin

    Corresponding author
    • Laboratory of Cellular Dynamics, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Search for more papers by this author
    • Iwona Ziomkiewicz and Anastasia Loman contributed equally to the work.


Correspondence to: Donna J. Arndt-Jovin, Laboratory of Cellular Dynamics, Max Planck Institute for Biophysical Chemistry, 37077 Göttingen, Germany. E-mail: djovin@gwdg.de

Abstract

We have revealed a reorientation of ectodomain I of the epidermal growth factor receptor (EGFR; ErbB1; Her1) in living CHO cells expressing the receptor, upon binding of the native ligand EGF. The state of the unliganded, nonactivated EGFR was compared to that exhibited after ligand addition in the presence of a kinase inhibitor that prevents endocytosis but does not interfere with binding or the ensuing conformational rearrangements. To perform these experiments, we constructed a transgene EGFR with an acyl carrier protein sequence between the signal peptide and the EGFR mature protein sequence. This protein, which behaves similarly to wild-type EGFR with respect to EGF binding, activation, and internalization, can be labeled at a specific serine in the acyl carrier tag with a fluorophore incorporated into a 4′-phosphopantetheine (P-pant) conjugate transferred enzymatically from the corresponding CoA derivative. By measuring Förster resonance energy transfer between a molecule of Atto390 covalently attached to EGFR in this manner and a novel lipid probe NR12S distributed exclusively in the outer leaflet of the plasma membrane, we determined the apparent relative separation of ectodomain I from the membrane under nonactivating and activating conditions. The data indicate that the unliganded domain I of the EGFR receptor is situated much closer to the membrane before EGF addition, supporting the model of a self-inhibited configuration of the inactive receptor in quiescent cells. © 2013 International Society for Advancement of Cytometry

The X-ray crystallographic structures of the extracellular domain of the epidermal growth factor receptor (EGFR) have been determined in both the liganded [1, 2] and unliganded [3] states. In the structures with bound EGF or TGFα, the ligand is associated with an extended region comprising domains I and III, and a homodimeric structure is generated by interaction of β-hairpins protruding from the two domains II of the constituent monomers, confirming an earlier model of Lemmon et al. [4]. This extended structure was also seen in the extracellular domain of the ErbB2 member of the ErbB family [5]. ErbB2 lacks a known ligand but features a potent kinase domain requiring activation by its preferred heterodimerization partners EGFR and ErbB3. In contrast, the unliganded extracellular domain of EGFR crystallized in a self-inhibitory structure in which the domain II dimerization arm is completely suppressed by intramolecular interactions with domain IV [3]. This structure has been presumed to be the form assumed by the inactive native receptor in the plasma membrane. A number of in vivo studies featuring mutated EGFR [6], antibody binding [7], and co-crystal structures of mAbs with native EGFR [8, 9] have confirmed the existence of a configuration in which the EGF-binding face of domain III is exposed. Lemmon [10] has suggested that EGFR be considered an allosteric enzyme, such that a complete understanding of the receptor requires combining cellular and structural studies. However, a recent study has challenged the existence of the inhibited conformation of the receptor in live cells. A chimeric full-length EGFR with YFP fused to its amino-terminus was reported to exclusively exist in the extended conformation both in the presence and absence of EGF [11].

We have addressed these issues by determining the orientation of the ectodomain of the full-length EGFR in quiescent cells compared to that exhibited after the addition of EGF. For these purposes, we constructed a new chimeric EGFR with an acyl carrier protein (ACP) tag fused to the wild-type amino terminus consisting of a three-helix bundle of 78 amino acids in which the serine in position 36 at the end of helix 2 served as the target for the enzymatic transfer by specific phosphopantetheine transferases (PPTases) of a fluorescent probe from the corresponding thiol-substituted CoA substrate. The ACP tag is a small entity of ≤7 kDa and is very versatile in terms of the labeling moiety, which can be a fluorophore, a quantum dot, or an affinity probe [12, 13].

Stably transfected cell lines were established and used in this study. The mutated EGFR protein was expressed well in mammalian cells and inserted into the extracellular membrane. It exhibited the properties of wild-type EGFR; that is, the kinase was dormant in starved cells but the receptor-bound EGF was thereby activated normally, showing the kinetics of downstream signaling and endocytosis characteristic of native EGFR (see Supporting Information Figs. S1 and S2).

Figure 1.

Schematic of the experimental design. Stick representations of full-length EGFR in the unliganded and inhibited conformation (left) and liganded dimer (right); adapted from illustrations in Ferguson [15]. The blue box on the ACP tag attached to domain I denotes Atto390. The “thunderbolts” in the cell membrane denote the acceptor NR12S. The downward arrow denotes the 740 nm two-photon laser excitation pulse. The Atto390 on the left shows quenching (lighter blue) due to resonance energy transfer to the nearest NR12S acceptor that shows increased fluorescence (red) compared with the Atto390 on the right.

Figure 2.

Molecular dynamics estimations of the distance of the Atto390-ACP-tagged EGFR to the C-terminus of domain IV in the unactivated, inhibited, and EGF ligand-bound conformations. A: Model of the structure of the ectodomain of the self-inhibited, resting, chimeric ACP-EGFR (1NQL) [3] with Atto390 attached covalently to serine 36 of the ACP moiety, showing its calculated distance to the C-terminus of domain IV after energy minimization and MD simulation. ACP, cyan; domain I 1–165, blue; domain II 166–309, green; domain III 310–481, yellow; domain IV 482–614, red with C-terminus orange. B: Model of the structure of the ectodomain of the extended, ligand-bound, chimeric ACP-EGFR (3NJP) [17] with Atto390 attached covalently to serine 36 of the ACP moiety, showing its calculated distance to the C-terminus of domain IV after energy minimization and MD simulation. Same color coding as in A.

The mutated EGFR protein was labeled with a donor fluorophore Atto390, capable of excitation by a multiphoton fs Ti–Sapphire laser, and images were acquired by the TCSPC technique. It was anticipated that fluorescence of the donor on the ACP tag would be reduced by FRET to the NR12S lipid probe localized in the outer leaflet of the cell membrane [14]. Based on molecular modeling predicting donor-membrane distances varying from ∼6 to 12 nm (see Results section) and an Ro of 5.4 nm, it was anticipated that the FRET efficiency would vary significantly with the absence or presence of the EGF ligand, that is, conditions representing the quiescent or ligand-bound state, respectively. The experimental strategy is summarized in Figure 1.

Our data demonstrate significant differences in the distance of the amino terminus of EGFR to the plasma membrane in the unliganded and liganded states, supporting the model of a self-inhibited conformation of the unliganded receptor in the quiescent state.

Materials and Methods

Molecular Modeling

The models of EGFR with an attached ACP tag in the putative inhibited and activated conformations were constructed with PyMOL [16]. The structures were energy-minimized and then subjected to short (10 ns) molecular dynamics (MD) runs to allow for a reorientation of the ACP with respect to the EGFR. The minimization and MD simulations were carried out on fully hydrated systems (TIP3P water) and 150 mM NaCl bulk salt concentration (Fig. 2). The amber99sb-ildn force-field [18] was used with GROMACS 4.55 [19]. After these procedures, the Atto390 probe with its linker was attached to SER36 of the ACP (Fig. 2).

Figure 3.

FRET donor and acceptor spectra and lifetime analysis. A: Donor and acceptor chemical structures and normalized excitation spectra (solid lines) and emission spectra (discontinuous lines). The fluorophore moieties are enclosed by blue rectangles. The calculation of inline image was performed [20] with the following parameters: co, 8.8·× 10−28; J, 1.54·× 1032; n, 1.333; inline image, 2/3; and inline image, 0.9 (manufacturer's data). B: Instrument response function (IRF). Second harmonic signal from KH2PO4 microcrystals; blue, data points; red, fitted function. C: TCSPC trace for the single pixel 2 (npix2) in the intensity collected data for this file in blue with superposition of a single component fit in red. D: TCSPC trace (blue) and two-component fit (red) for the lowest of the five intensity bins of the pixel values (see the full analysis information for this bin designated with the red arrows in E). E: Table of fit parameters for the five bins. See text for details. Errors are standard errors of the mean values (SEM), propagated according to the respective parameters.

Chimeric ACP-EGFR Protein and Expression in Cells

We constructed a chimeric EGFR with an amino terminal acyl carrier protein tag [22] attached to the amino terminus of the mature human EGFR. The DNA sequence for the 78 amino acid sequence of the ACP tag was extracted from the AGA-plasmid (kind gift of N. Johnsson) by PCR and inserted between the DNA for the signal sequence and the amino terminus by creating a new Nhe1 restriction enzyme cleavage site using site-directed mutagenesis of wild-type EGFR in pcDNA3 or a modified pcDNA3.1zeo-plasmid. Stably transfected CHO cell lines were selected by antibiotic resistance and single-cell cloning. The serine that accepts the 4'-phosphopantetheine (P-pant) conjugate occurs at residue 36 in the sequence of the mature expressed protein. The mean expression of the chimeric protein in the cell line used in this study was 1.35 ×·105 as determined by flow cytometry using a QiFIT DAKO kit and the monoclonal antibody 528 specific for the EGFR.

Synthesis of Atto390-CoA

Synthesis was carried out as described by Vivero-Pol et al. [13] by reaction of Atto390 maleimide (Atto-Tec GmbH) with the cysteine of CoA and purification by reverse-phase HPLC C18 chromatography.

Labeling of Cell Surface Expressed ACP–EGFR with Atto390 and of the Cellular Membrane with NR12S

Cells expressing the ACP–EGFR were cultured in 35-mm microdishes (Ibidi GmbH, cat. no. 81156) in DMEM (Life Technologies) supplemented with 10% fetal calf serum. Before measurements, the cells were brought into the resting state by serum starvation for at least 4 h. Enzymatic labeling of the ACP–EGFR on the plasma membrane was carried out in the dark for 10–20 min at room temperature in Tyrode's buffer supplemented with 10 mM MgCl2 by the Sfp PPTase from Bacillus subtilis and the substrate, 2 µM Atto390-CoA. Cells were washed several times with Tyrode's buffer. For determination of Förster resonance energy transfer (FRET), cells were labeled for an additional 5 min with the membrane probe NR12S [14] serving as the FRET acceptor, using a ≥500-fold fresh dilution of the dye from a DMSO stock solution. The dishes were filled with Tyrode's buffer supplemented with 2 mM glucose and 5 mg/ml BSA and inverted on the upright microscope stage, such that the 60× objective directly faced the chamber membrane surface with the adherent cells. Experiments were conducted at room temperature. For experiments on activated EGFR, the cells were incubated for 30 min before receptor labeling with 2 µM of the EGFR-specific kinase inhibitor PD153035 (Calbiochem). The inhibitor was present throughout the process of dye labeling and subsequent measurement. The human-recombinant EGF (Peprotech) was added to a concentration of 36 nM 3 min before the initiation of the measurements.

TCSPC Lifetime Image Acquisition

The excitation source was a multiphoton Ti:Sapphire fs laser (Chameleon Ultra, Coherent) tuned to 740 nm and pulsed at 80 MHz. Measurements were carried out on an upright Olympus scanning confocal microscope (Fluoview 1000) using a 60× 1.35 NA UPlanSApo oil immersion objective. The fluorescence signal at 480 nm (20 nm HW) was acquired with a SPAD detector (Micro Photon Devices), using the PicoHarp 300 TCSPC acquisition system of PicoQuant GmbH. Images were acquired with a scan speed of 20 µs/pixel at 256 × 256 pixel (x, y) resolution and 16-ps time resolution.

Lifetime Image Analysis

Analysis software was developed in Mathematica versions 7/8/9 (Wolfram Research). The program (ReadPicoQuant_bin.nb) reads and interprets bin files generated by the PicoQuant SymPhoTime program. The instrument response function (IRF) was obtained from the second harmonic-generated signals of KH2PO4 microcrystals excited at 960 nm and detected in the 480 nm (20 HW) emission window. Pixel signals were sorted according to intensity, and a mean response, computed from the top 392 levels, was fit to a seven-parameter (plus time) analytical function (pulse) consisting of a Gaussian distribution function plus an integrated Gaussian distribution function times an exponential decay (see Fig. 3B and Supporting Information).

In the case of the fluorescence images, an ROI was defined delineating the membrane, and the signals from the sorted pixels were fit to one- or two exponential decays, each component of which was characterized by an amplitude (counts) times a signal function, the convolution of the analytical IRF function with a normalized exponential decay [Eq. [1], Supporting Information]. The only free parameters in signal are the lifetime τ and a corresponding amplitude, all other parameters being derived from the IRF analysis. The advantage of signal is that it eliminates noise originating from deconvolution with an experimental IRF and fits the entire (ascending as well as descending) fluorescence time course. Despite the very asymmetric forcing function (IRF) of the TCSPC system, signal yields reproducible two-component analyses, even in the presence of short lifetime components that render tail-fitting routines unfeasible.

display math(1)

Because of the noise content, single pixel decays could only be fit with a single component (see example in Fig. 3C). In the case of the more precise two-component analyses of pixel data ordered and binned into groups (generally five, each with 100–1500 members) according to peak signal magnitude (see examples in Figs. 3D and 3E), a correction was applied to compensate for the data wraparound introduced by the 80 MHz repetition frequency. This procedure was implemented by adding to each given component the same component mirrored to a t + 1 period (12.5 ns). In most cases, temporal channels were merged (generally by a factor of 5×), providing a major calculation speedup with no significant change in fit parameters. The derived amplitudes and lifetimes were subjected to examination and interpretation based on the corresponding images and two-dimensional histograms.

FRET Formalism

We assume that the donor–acceptor relationship applicable to the system under study corresponds reasonably to the treatment of Bastiaens et al. [21] of a single donor separated by a given distance (along the normal), rDm (subscript: donor-membrane), from a plane of acceptors with surface density σ. In the form expressed in Eq. [2], the index i denotes the different experimental conditions and molecular configurations that can be adopted, kt,i is the rate constant for energy transfer, c0 is a constant [20], Ei is the FRET efficiency, inline image is the fluorescence lifetime of the donor in the absence of the acceptor, and the Förster transfer distance inline image is defined in terms of the canonical (rapid rotational redistribution) value (2/3) for the orientation factor inline image. The latter assumption introduces little error in the estimations of distances, particularly in the relative values [Eq. [3]], in the event that at least one partner is mobile [21]. The estimated inline image for the Atto390-NR12S pair is 5.4 nm (Fig. 3A). We favor the FRET expression in terms of the transfer rate rather than the transfer efficiency E because of its more direct functional relationship to the quantities of greatest interest in this study, rDm and σ [20]. Noteworthy is the inverse fourth-order dependence of the transfer rate constant on distance as opposed to the sixth-order dependence for an isolated donor–acceptor pair; kt is also proportional to σ, the acceptor density in the plasma membrane, for which the assumption of a uniform planar disposition is a reasonable simplification. For example, the effect of plasma membrane curvature on FRET has been shown to be negligible [23].

display math(2)

Results

Molecular Modeling of the Atto390 Fluorophore Covalently Coupled to the Acyl Carrier Protein–EGFR Ectodomain in the Self-Inhibited and in the Liganded Conformations

Molecular model building was performed by combining the coordinates from the PDB structures 3NJP [17] of the EGFR liganded extracellular receptor domains (I–IV), 1NQL [3] of the EGFR unliganded extracellular receptor domains (I–IV), and the NMR coordinates from the PDB structure 1ACP [24] for the ACP motif. In the chimeric protein, the latter is attached to the amino terminal amino acid (shown as blue spheres in the diagrams) of the full-length EGFR sequence. According to the structures optimized using GROMACS 4.55 [19] (Figs. 2A and 2B), the Atto390 fluorophore (Fig. 3A) attached to serine 36 in the ACP motif would be located ∼60 and ∼120 Å, respectively, from the C terminus of domain IV (shown as gold spheres), which connects to the transmembrane segment and cytoplasmic domains (not shown) in both the presumptive unliganded and liganded states. The depicted structures are snapshots of an ensemble of possible conformations of the attached tag. Exact distance estimations are not feasible, because rather short 10 ns MD simulations were carried out. However, the approximate twofold difference in the distance of the dye molecule to the outer leaflet of the membrane, where the NR12S dye is located, would be expected to result in a large difference in the donor lifetime because of the fourth power distance dependence of FRET [Eq. [2]].

Labeling of the Chimeric ACP–EGFR with Donor Atto390 and the External Leaflet of the Cellular Membrane with Acceptor NR12S

The ACP–EGFR protein was expressed well in mammalian cells and inserted into the extracellular membrane. It exhibited the properties of wild-type EGFR; that is, the receptor bound EGF and was transphosphorylated as well as showing the kinetics of downstream signaling and endocytosis characteristic of native EGFR (see Supporting Information Figs. S1 and S2). Labeling of the ACP tag by the donor fluorophore Atto390-CoA by Sfp PPTase proceeded as described in the Materials and Methods section.

Lipid soluble fluorescence dyes can be used to externally label the plasma membranes, but most exhibit rapid flip-flop between the outer and inner leaflets and are internalized rapidly, distributing to all intracellular membrane compartments. Recently, Kucherak et al. [14] described a derivative of Nile Red, NR12S, conjugated with a zwitterionic group and a long hydrophobic chain (Fig. 3A). This compound selectively stains the outer leaflet of lipid bilayers and exhibits negligible flip-flop, leading to exceptionally long retention time at plasma membranes of mammalian cells [14, 25]. The spectral properties of NR12S incorporated into lipid bilayers are complex, with solvatochromic shifts in both absorption and emission as a function of lipid composition and order [14]. In the cellular context, the absorption spectrum has a peak in the 525–530 nm range and the fluorescence spectrum at 585–605 nm. NR12S was selected as a good FRET acceptor match for the Atto390 donor fluorophore attached to EGFR (see below). Atto390 exhibits a large Stokes shift, such that its excitation maximum (390 nm) lies at a minimum in NR12S absorption while its emission (peak at 479 nm) has a good overlap with the absorption spectrum of NR12S in lipid membranes with a calculated Ro of 5.4 nm (Fig. 3A).

TCSPC Lifetime Measurements of FRET in Living Mammalian Cells

Images of cells labeled with the Atto390 donor in the absence or presence of the NR12S acceptor showed a strong 480 ± 20 nm signal localized to the plasma membrane upon two-photon excitation at 740 nm. FRET occurring in the presence of the NR12S acceptor was assessed by the reduction of the donor lifetime [26], because the large excess of acceptor emission excited directly in all regions of the plasma membrane precluded methods based on sensitized emission. Image pixels were sorted by intensity and subjected to a single component decay analysis, yielding lifetimes of 4.6 ± 0.3 ns for the top 100 intensity values under all conditions (Fig. 3C). Cells lacking the Atto390 label exhibited low-level background signals originating from generalized intracellular locations and with lifetime fits of <1 ns. This component is primarily attributable to NADH [27] and was minimized by applying a mask restricting the analyzed pixels primarily to the plasma membrane. The mask was double-sided and could also be used to limit the intensity selection to any arbitrary region, for example, so as to exclude the highest values. Spatial binning was not used in order to minimize the influence of ambiguous edge pixels.

After binning the foreground pixels into groups of 100–1500, it was possible to perform reproducible lifetime determinations using the formalism accounting for two components as well as the perturbation arising from the short measurement period of 12.5 ns [Eq. [1] and Fig. 3D]. The analysis program provided statistical measures for all the fit parameters [two amplitudes, two lifetimes, two intensities (relative amplitude × lifetime), and mean (amplitude-weighed) lifetime] (Fig. 3E), from which two-dimensional histograms and images showing the spatial distribution of the computed lifetimes were generated (see examples in Figs. 4-7 and Table 1). The fluorescence decay of the Atto390 probe was perceived as a single major component ( inline image) of >5 ns accounting for 87–98% of the signal intensity, and the presence of FRET was manifested by changes in inline image. The lifetime of the donor reproduces that supplied by the manufacturer as well as our own determinations in another study of Atto390 attached to nucleic acids, which yielded values of 6.0 and 1.6 ns (93% and 7% intensity, respectively). Neither the single component fits, as indicated earlier, nor the amplitude-weighted mean lifetimes derived from two-component fits, generally used as a FRET measure in imaging applications, were useful due to the presence of the variable shorter ∼1 ns ( inline image) component unaffected appreciably by FRET.

Figure 4.

Gallery of images of cells labeled only with donor, Atto390-ACP-EGFR. Left, intensity images of the donor emission in the selected ROI; middle, masked intensity image; right, inline image values corresponding to the masked pixels. The LUT at the left applies to the rescaled peak intensity for the left and middle images and for the given range of lifetime values (right image).

Table 1. Lifetime analyses and FRET calculations
Sample[NR12S] (μM) inline image(ns) inline imageE
  1. Error: standard error of the mean (SEM). For other definitions, see Eq. [1].

Atto390-ACP-EGFR05.26 ± 0.06 
0.24.97 ± 0.05 0.06 ± 0.01
0.54.81 ± 0.031.8±1.10.09 ± 0.01
1.04.81 ± 0.01 0.09 ± 0.01
Atto390-ACP-EGFR + EGF + PD0.25.22 ± 0.04 
0.55.10 ± 0.03 0.03 ± 0.01
Atto390-ACP-EGFR + PD05.17 ± 0.12  

Measurements on Starved Cells

Figure 4 features images from different experiments of three cells labeled with the donor Atto390. The color-coded images are of the total intensity, the intensity processed with a binary mask, and the corresponding determinations of inline image obtained by binning the pixels into five groups and fitting by two-component analysis. The global mean ± SEM of inline image from these and other data (n = 9 images, each of 1–2 cells, with 500–7500 analyzed pixels) was 5.26 ± 0.06 ns (see Table 1 and Fig. 7). The distribution of lifetimes in any given cell was fairly narrow and lacked an apparent correlation with intensity. Measurement of the donor in the presence of the kinase inhibitor PD153035 without added acceptor also yielded similar lifetimes (Table 1). In a limited set of measurements, fixation of the Atto390-labeled cells (3.7% PFA, 10–15 min at room temperature) was shown not to have had a discernible effect on the lifetimes.

Quiescent Atto390-labeled living cells were exposed to increasing concentrations (0.2–1 µM) of the NR12S acceptor probe. A systematic and consistent diminution of inline image was observed (Table 1, Fig. 7). It was determined that 0.5 and 1 µM NR12S led to equivalent values, and therefore, the majority of experiments were conducted with 0.5 µM acceptor. The global mean ± SEM of inline image was 4.81 ± 0.03 ns (n = 15), corresponding to a decrease of 9%. Images of signals from cells with donor plus 0.5 µM acceptor and the calculated inline image lifetime are shown in Figure 5. The inline image values for 0.2 and 1 µM NR12S were 4.97 ± 0.05 ns (n = 3) and 4.81 ± 0.01 ns (n = 2), respectively (Fig. 7). Equation [2] predicts a linear dependency of the transfer rate on acceptor density. A more detailed assessment (additional acceptor concentrations) was not feasible in the context of these experiments. Furthermore, elucidation of the relationship between membrane loading and external dye concentration, the dependence on incubation conditions, and of the uniformity of probe distribution and properties, particularly in the vicinity of the receptor, would constitute a separate investigation.

Figure 5.

Gallery of images of cells labeled with donor and acceptor (0.5 µM NR12S). Details as in Figure 4.

Measurements After Addition of EGF

The ligand EGF was added in the presence of a specific EGFR kinase inhibitor PD153035. The latter does not affect EGF binding but prevents endocytosis, thereby retaining the receptor in its plasma membrane environment. Addition of saturating concentrations of EGF increased the donor lifetime to inline image values (5.10 ± 0.03, n = 12), close to those obtained in the absence of acceptor (3.0% lower). Images of some of these cells are shown in Figure 6.

Figure 6.

Gallery of images of cells labeled with donor and acceptor (0.5 µM NR12S) in the presence of EGF and kinase inhibitor PD153035. Details as in Figure 4.

Figure 7.

Fluorescence lifetimes ( inline image) from four independent experiments and different conditions. The abscissa denotes the different experimental conditions: the concentration of acceptor present and the presence of EGF where applicable. The mean inline image values are indicated by horizontal lines for each condition. See text for details.

Interpretation of the FRET Data in Terms of Distances

The results of several sets of experiments conducted under the various conditions are presented in Table 1 and Figure 7. The data show a consistently shorter donor lifetime for the starved cells in the presence of acceptor than those measured after binding EGF. One can use Eq. [2] in two ways to estimate consistency with the distance estimations from the molecular modeling (Fig. 2). We first apply the equation to each condition of interest (a: donor; b: donor + acceptor; c: donor + acceptor + EGF) and evaluate the following relationship, in which common parameters ( inline image) conveniently cancel in the ratio.

display math(3)

Introduction of the respective values of the liganded and unliganded forms of the EGFR into Eq. [3] yielded a value of 1.3 ± 0.2 (SEM), which can be compared to the (single) estimation of 11.2/6.2 = 1.8 from molecular modeling (Fig. 1). The agreement is not unreasonable in view of the conceptual and experimental uncertainties associated with these measurements and simulations. Thus, the value of 1.8 would be reduced to 1.5–1.7 by adding the finite distance between the C-terminus of domain IV and the NR12S probe inserted into the outer leaflet of the plasma membrane to both the numerator and denominator of the molecular modeling ratio; we estimate this contribution as ∼1–3 nm [28].

A second application of Eq. [2] is in the comparison of the most reliable FRET data of condition b with those obtained under condition c at the same acceptor concentration. A range of rDm,b 6.0–6.4 nm was calculated by using the mean inline image values and specifying an acceptor density σ of 0.003–0.004 probe molecules/nm2. The same range of σ, applied to condition c, that is, after addition of EGF, would correspond to an rDm,c of >7.8 nm. Reliable quantitative determinations of membrane probes applied externally to cells are not available in the literature. One study using bulk probe concentrations of 0.2–7.5 µM yielded estimates for σ of 0.001–0.01 [29], values consistent with our calculations. Table 1 also provides an estimate of the transfer rate and energy transfer efficiencies.

The above methods of analysis provide convincing evidence for a significant extension of the EGFR after addition of EGF, such that the distance of the N-terminus from the membrane increases, in conformity with the model featured in Figures 1 and 2.

Discussion

The FRET–FLIM determinations demonstrate that the N-terminus of the unliganded EGFR in the quiescent living cell is situated much closer to the membrane than after EGF addition. This result supports the model according to which the inactive receptor in quiescent cells is predominantly in the self-inhibited configuration [7, 9, 30] shown in Figure 2A. Starving the cells as well as measurements at room temperature favor the quiescent structure. In contrast, the X-ray crystallographic structure of the liganded EGFR features monomers extended and dimerized in a back-to-back configuration [1, 2], such that the distance of the amino terminus from the cell membrane (Fig. 2B) would be approximately double. The molecular modeling is compatible with these interpretations in that it predicts that the extended liganded EGFR dimer would have a distance of the ACP–EGFR domain I to the cell membrane considerably greater than the inline image for the Atto 390-NR12S donor–acceptor pair of 5.4 nm and should thus show little energy transfer, whereas the inhibited structure would bring the dye pair close to the Förster transfer distance. Our quantitative treatment of the FRET–FLIM data was based on a formalism applicable to transfer from a donor probe located in the ectodomain of a transmembrane protein to an array of acceptor probes integrated into the supporting lipid bilayer [21].

The structures shown in Figure 2 are snapshots of the presumed configurations adopted by the EGFR in the living cell, inasmuch as the MD simulations were initiated from the X-ray self-inhibited or liganded species, respectively. Although the interdomain contacts of domains II and IV of the self-inhibited structure remained and the molecule did not extend during the MD simulations, we assume that in living cells, there exists flexibility in both structures, including a “breathing” of the inhibited form. As mentioned earlier, the actual distance between the donor and acceptor is probably larger than that calculated from the molecular modeling, because a seven amino acid linker connects the C-terminus of domain IV to the transmembrane segment [17] and the acceptor dye moiety is presumably immersed in the membrane. Using the original diffraction data set of Ogiso et al. [2], Lu et al. [17] calculated a refined annealing-composite omit map to determine the position of domain IV in the liganded dimer. The structure of domain IV could be superposed by bending and displacing the C-termini by 10 Å from the monomer [3] to dimer structures [17]. Disulfide crosslinking studies of cells expressing single cysteine replacements established that the seven amino acid linker between the C terminus (residue 614) of domain IV, and the transmembrane segment is highly flexible but has no preferential pairing upon EGF binding [17]. Similar disulfide crosslinking with substitutions of the amino acids in the 23-residue transmembrane segment, however, exhibited a strong peak at residue 624 and a weaker one at 627 only upon EGF binding, indicating an association of the transmembrane domains of the dimer only in the first two α-helical turns.

The fact that the linker region is flexible allows other orientations of EGFR, such that inactive, unliganded dimers might form as proposed by Jura et al. [31], or asymmetric dimers might form as seen with the invertebrate Drosophila EGFR [32]. The latter viewpoint has been championed by the group of Martin-Fernandez [33-35], which has focused on biophysical measurements to elucidate the structure of the “high-affinity” EGFR implicated by Scatchard EGF-binding plots showing upward curvature (negative cooperativity) at very low-EGF concentrations. This group has presented evidence from FRET/FLIM measurements with a labeled EGF ligand (<1 nM) as donor and a membrane carbocyanine dye as acceptor for the existence of a small cohort (10–15%) of “high-affinity” dimers, which would be asymmetric and lie on (parallel to) the cell membrane. Unfortunately, the use of EGF as the donor instead of a label attached to the receptor as in the present study, does not permit a useful comparison of the inhibited monomer conformation from the extended dimer configurations inasmuch as EGF interacts with the open domain III in the inhibited conformation and is thus situated at a similar distance from the membrane in both forms. Our measurements, performed exclusively with saturating concentrations of EGF, gave no indication of a substantial contribution from a structure of the liganded receptor with the N-terminus situated close to the membrane.

A previous publication reported little FRET between an amino terminal fused YFP–EGFR chimera and concluded that the EGFR receptor in the cell membrane exists in an extended form and probably as a constitutive dimer [11]. This transgene, however, shows spontaneous activation of the receptor in the absence of EGF, a phenomenon that does not occur in starved cells expressing wild-type receptor, suggesting that the YFP-chimera may have been dimerizing via the YFP moiety and not adopting the normal configuration of the normal receptor in the membrane. It is well known that YFP without the A206K mutation undergoes spontaneous dimerization [36]. However, this mutation was not used in the Kozer et al. [11] study. In addition, the Rh-DHPE (LissamineTM rhodamine B 1,2-dihexadecanoyl-sn-glycero-3-phospho-ethanol-amine) acceptor used by these authors was not restricted to the outer leaflet of the cell membrane. Finally, Kozer et al. [11] used a hematopoietic cell-line BaF/3 for their biophysical studies. Although this cell line lacks endogenous ErbB proteins, it is a very different than the epithelial cells that normally express the ErbB family. We assume that both the cell type and the large-added protein domain led to significant changes in the disposition of the EGFR ectodomain compared to other cells expressing EGFR. Only the studies of such hematopoietic cell lines have led to reports of a large fraction of dimers of EGFR in resting cells with expression levels of 50,000 receptors and of signaling being initiated from tetramers [37]. In contrast, studies of CHO and HeLa cell lines have shown predominantly monomeric EGFR receptors in cells expressing <100,000 receptors by a variety of measurements including fluorescence correlation spectroscopy (FCS) [38], number and brightness (N&B) [39], and single particle tracking (SPT) [40]. Unliganded dimer formation is <30% even in cells with receptor numbers of 5 ×·105 [39]. SPT measurements have demonstrated similar diffusion constants of unliganded and liganded receptors even in A431 cells with 2 ×·106 receptors, ruling out a predominance of preformed dimers and tetramers as preferential signaling platforms. As we discuss below, both the lipid composition of the BaF/3 cells and the promotion of dimer formation by YFP interactions probably account for the failure of Kozer et al. [11] to detect the inhibited conformation.

Assuming that the receptor is in a dynamic equilibrium between the extended (dimeric) and inhibited conformations, interactions preferential for one or the other conformation will perturb the fractional distribution. The high-lateral mobility of the EGFR, for example, determined by Lidke et al. [41], by FRAP, will contribute to population redistribution. From N&B determinations, Nagy et al. [39] inferred a greater fractional population of unliganded dimers in cells expressing 500,000 receptors than in those with 50,000 receptors, and Chung et al. [42] observed more dimers in cells with high-receptor densities using SPT. Saffarain et al. [38] showed by FCS that depletion of cholesterol from cells increased the number of unliganded dimers in living cells. Recent studies have also revealed that certain lipids have a profound effect on EGFR structure and activation [43]. EGFR binds seven peptide growth factors, which elicit distinct physiological responses [44]. It is possible that the conformational rearrangements induced by these natural ligands are somewhat different than those induced by EGF, suggesting that studies of these complexes in living cells by FRET–FLIM measurements may be informative.

The Bastiaens group has studied the key role of phosphatases in the deactivation of spontaneous signaling by EGFR from the plasma membrane [45, 46]. Their data suggest that spontaneous dimer formation can lead easily to transphosphorylation, constituting an activation mechanism operating in the absence of ligand. EGFR is overexpressed or mutated in many cancers, leading to increased activation and consequent unhindered growth [47, 48]. A more striking example of the importance of the self-inhibitory structure is the constitutively activated mutant EGFRvIII found in many gliomas [49, 50], which has a deletion of residues 6–273 that eliminates parts of domains I and II. This mutant receptor cannot adopt the inhibited structure and thus may allow the transmembrane and kinase domains to more easily assume their activated conformations. It is important that the majority of the EGFR in normal epithelial cells be maintained in the inhibited state and that spontaneous dimer formation be controlled so as to restrain aberrant activation of the cell cycle and proliferation. Our data support the model of the quiescent monomer receptor in which part of the inhibition to self-activation is provided by a conformation of the receptor that features interaction of domain II with domain IV and brings domain I close to albeit not directly at the cell membrane.

During the revision of this report, a study appeared featuring extensive modeling by MD of various forms of the EGFR in a membrane environment [51]. The authors concluded that the ectodomains of the monomeric and unliganded dimeric forms are intimately associated with the membrane, with only the active (liganded) dimer in an extended (“upright”) configuration. Our data and conclusions do not support such a view, emphasizing the importance of as yet not fully understood structural and microenvironmental features of these complex molecules located in their native cellular milieu.

Acknowledgments

I.Z. was a student of the Faculty of Biology of the Adam Mickiewicz University, Poznan, Poland, supported by an Erasmus EU exchange fellowship with the Georg-August-University of Göttingen. A.L. was supported by the German Federal Ministry of Education and Research (BMBF). The authors thank F. Wouters for providing the multiphoton laser system. D.J.A-J. and T.M.J. conceived of the experiments; C.F and D.J.A-J. made the EGFR constructs and isolated the cell lines; D.J.A-J. synthesized and purified the Atto390 CoA compound; I.Z and A.L. performed the measurements; R.K. performed the model building; A.S.K. provided the NR12S; G.B. helped with the alignment of the laser and guided the TCSPC measurements; T.M.J. developed the analysis software and analyzed the data; D.J.A-J. and T.M.J. wrote the manuscript. The authors declare no conflict of interest.

Ancillary