EGFR forms ligand-independent oligomers that are distinct from the active state

The human epidermal growth factor receptor (EGFR/ERBB1) is a receptor tyrosine kinase (RTK) that forms activated oligomers in response to ligand. Much evidence indicates that EGFR/ ERBB1 also forms oligomers in the absence of ligand, but the structure and physiological role of these ligand-independent oligomers remain unclear. To examine these features, we use fluorescence microscopy to measure the oligomer stability and FRET efficiency for homo- and hetero-oligomers of fluorescent protein-labeled forms of EGFR and its paralog, human epidermal growth factor receptor 2 (HER2/ERBB2) in vesicles derived from mammalian cell membranes. We observe that both receptors form ligand-independent oligomers at physiological plasma membrane concentrations. Mutations introduced in the kinase region at the active state asymmetric kinase dimer interface do not affect the stability of ligand-independent EGFR oligomers. These results indicate that ligand-independ-ent EGFR oligomers form using interactions that are distinct from the EGFR active state.

Human epidermal growth factor receptor (EGFR/ERBB1) and its paralogs human epidermal growth factor receptor 2 (HER2/ERBB2), HER3/ERBB3, and HER4/ERBB4, collectively known as ERBBs, are receptor tyrosine kinases (RTKs) that are essential for normal growth and development (1). Abnormal activation of both EGFR and HER2 is associated with multiple human cancer types, and each is the target of several anticancer agents (2). ERBBs are Type I integral membrane proteins composed of an extracellular domain (ECD) that is made up of four distinct subdomains, an a-helical transmembrane domain (TMD), and an intracellular domain comprising a juxtamembrane region, a tyrosine kinase domain, and a nonglobular ;230 amino acid C-terminal tail. The canonical mechanism by which ERBBs are thought to act is ligand-dependent oligomerization, which results in stimulation of the intracellular kinase, trans-autophosphorylation, recruitment of effector proteins, and initiation of intracellular signaling cascades (1,(3)(4)(5).
Mutations or deletions in EGFR and HER2 result in ligandindependent receptor activation, indicating that in addition to being activated in the presence of ligand, ERBBs are actively inhibited in the absence of ligand (25)(26)(27)(28)(29)(30). The mechanism for this autoinhibition remains incompletely understood, however. In contrast, much is known about how ligand binding promotes ERBB activation (31,32). In the absence of ligand, the EGFR, ERBB3, and ERBB4 ECDs adopt a tethered conformation that buries an extended b-hairpin loop (33)(34)(35)(36). Binding of ligand to these ECDs stabilizes an extended conformation in which an extended b-hairpin (termed the dimerization arm) is exposed and mediates formation of an active receptor dimer. In this dimer the intracellular kinase domains adopt an asymmetric dimer interaction in which the C-lobe of a donor kinase contacts the N-lobe of a receiver kinase, stabilizing the active conformation of the receiver kinase (37). Comparatively little is known about the structure and physiological role of either the TMD or the intracellular domain in the inactive state, although mechanisms by which the TMD may function in the inactive state have been proposed (35,38,39).
Here we report quantitative FRET measurements of homoand hetero-interactions between near full-length EGFR and HER2 in vesicles derived from plasma membranes. We confirm previous findings that EGFR and HER2 are in dynamic equilibrium between monomeric, homo-oligomeric and hetero-oligomeric forms at physiological receptor concentrations (40). Because we are unable to distinguish dimers from higher-order oligomers (n 2) by our method we refer to oligomers rather than dimers, although single-molecule photobleaching and fluorescence intensity fluctuation studies indicate that ligand-independent EGFR oligomers are predominantly dimeric (21,23,41). We observe that mutations in the asymmetric kinase interface do not preclude formation of ligand-independent EGFR oligomers, indicating that an alternate conformation to the active dimer exists within the ligand-independent oligomer ensemble.

Quantitative imaging FRET microscopy
To enable measurement of homo-and hetero-ERBB interactions in native cell membranes using FRET microscopy, human EGFR and HER2 variants labeled at their C termini with a fluorescent donor (EYFP) or acceptor (mCherry) protein were expressed in Chinese hamster ovary (CHO) cells. ERBBs have ;200 amino acid disordered regions at their C termini, which, along with 8 amino acid Gly-Ser linkers inserted between each ERBB and its fluorescent protein (fp) fusion partner, are presumed to mitigate effects of fluorescent dipole orientation on observed FRET (Fig. S1) (42). Expression levels of full-length forms of fp-tagged EGFR and HER2 proved too low to characterize interactions with confidence. EGFR and HER2 variants with C-terminal deletions were thus tested to identify minimal deletions that preserve activity and result in improved expression levels. Removal of 107 and 25 residues from the C termini of EGFR and HER2, respectively, generated the variants EGFR-D107-fp and HER2-D25-fp and resulted in proteins that express well and are functional as judged by anti-phosphotyrosine Western blots (Fig. 1A). EGFR variants with a 237 amino acid C-terminal deletion are able to elicit EGF-dependent cellular functions, indicating that the region deleted in EGFR-D107-fp is not essential for activation of downstream effectors (43,44).
Homo-and hetero-interactions between EGFR-D107-fp and HER2-D25-fp were measured using quantitative imaging FRET (QI-FRET) microscopy (45)(46)(47). Briefly, plasma membrane vesicles from CHO cells transiently transfected with fp-labeled EGFR and/or HER2 were generated by incubating live cells in a hyperosmotic solution (Fig. 1B) (48,49). Endogenous EGFR in vesicles derived from A431 cells retains EGF-dependent phosphorylation indicating that vesiculation does not impair ERBB function (Fig. 1C). The concentrations of donor-and acceptorlabeled ERBBs as well as the FRET efficiency between labeled ERBBs in vesicles were obtained from three fluorescence scans (Fig. 1D). Cells in the same transfection experiment display a wide range of receptor expression levels as judged by fluorescence intensity (Fig. 1B), which allows FRET efficiency to be determined as a function of receptor concentration. A wide range (roughly 10-fold) was observed for all variants tested, and the identity of the fluorescent protein (EYFP or mCherry) did not affect the variability in the observed concentration. Following correction of FRET signals for FRET arising from densitydependent proximity (known as proximity or bystander FRET) (50) (Figs. S2-S4) and for the relative fraction of donors and acceptors in each vesicle (45,46), apparent two-dimensional dissociation constants can be determined.
The oligomeric state of fluorescently labeled proteins affects the FRET signal (51). The FRET changes we observe fit a monomer-dimer model as well as monomer-dimer-tetramer or monomer-trimer models. We thus report fits to a monomerdimer equilibrium, although the receptors likely form higherorder oligomers in the presence of EGF (23,41). This fit depends on two parameters: the intrinsic FRET efficiency, Ẽ, which equals the FRET efficiency within a dimer, and the twodimensional dissociation constant, K d , in units of molecules/ mm 2 (45,46). In cases where the association is strong at all receptor concentrations, we assume that the intrinsic FRET value (Ẽ) is equal to the mean observed FRET value. As the presence of higher-order oligomers is likely (41), especially when ligand is present, we use the term oligomer throughout to denote complexes of two or possibly more receptors.

Near full-length EGFR forms ligand-independent oligomers at physiological plasma-membrane concentrations
The FRET efficiency for EGFR-D107-fp increased as a function of concentration and fit well to a monomer-dimer equilibrium model, with a two-dimensional K d of 156 molecules/mm 2 ( Fig. 2A and B) and a FRET efficiency value of 0.35 6 0.06 (Table 1), consistent with previous observations (23,24). Addition of EGF or substitution of the EGFR extracellular domain with a constitutively dimeric immunoglobulin Fc-domain results in constant FRET efficiency values of 0.29 6 0.03 and 0.29 6 0.02, respectively, as well as constitutive receptor phosphorylation ( Fig. 2E and F, and Fig. S5). The cancer-associated L858R mutation has been shown to promote EGFR oligomerization in the absence of ligand (52), and introduction of L858R into EGFR-D107-fp results in higher FRET efficiency values over all concentration ranges measured indicating increased self-association in the absence of ligand ( Fig. 2C and D). Observation of higher FRET efficiencies under conditions known to activate EGFR provides confidence that our observations reflect physiological behavior. The FRET efficiency differences become less apparent when the FRET data are binned over all observed receptor concentrations (Fig. 2B, D, and F), underscoring the need for measurements over a range of concentrations.

HER2 forms homo-and hetero-oligomers in the absence of ligand
FRET efficiency values for self-association of HER2-D25-fp and associations between HER2-D25-fp and EGFR-D107-fp were ;0.3 and varied little with concentration, consistent with formation of constitutive ligand-independent HER2 homooligomers and EGFR/HER2 hetero-oligomers ( Fig. 3 and Table  1). These results indicate HER2 self-associates more strongly than EGFR in the absence of ligand.

EGF-independent EGFR phosphorylation increases with increasing EGFR expression but has no measurable effect on STAT or ERK phosphorylation
Consistent with previous observations, ligand-independent EGFR oligomer formation increases at higher EGFR concentrations, in accordance with the law of mass action (Fig. 2) (23,24). To assess the activity of EGFR oligomers formed in the absence of ligand, we generated a panel of CHO cell lines stably expressing full-length EGFR with EYFP at its C terminus (EGFR-EYFP) and sorted cells to select cell lines expressing increasing EGFR-EYFP levels. The apparent EGFR-EYFP surface concentration of selected cell lines was measured by observing fluorescence in vesicles derived from each cell line. Despite extensive effort, we were unable to isolate a cell line stably expressing EGFR-EYFP at levels higher than ;200 receptors per mm 2 on average (Fig. 4A), which is somewhat lower than the 646/mm 2 reported for A431 cells, a cell line with the highest known EGFR expression level (53). This concentration is nevertheless similar to the value observed for the dissociation constant for EGFR-D107-fp ( Table 1) and indicates that a substantial fraction of EGFR-EYFP is likely to exist in oligomers in these cell lines.
Seven cell lines selected for increasing EGFR-EYFP expression levels were treated with EGF and analyzed by Western blotting for EGFR and phosphotyrosine (Fig. 4, A-C). In the absence of ligand, EGFR phosphorylation was not detectable at concentration levels below 100 molecules/mm 2 . Between 100 and 300 molecules/mm 2 , ligand-independent phosphorylation increased in a linear fashion. Stimulation with EGF resulted in a linear increase in phosphorylation across all measured expression levels (Fig. 4C). Notably, the EGFR-EYFP expression range over which EGF-independent phosphorylation begins to become detectable (100-300 molecules/mm 2 ) is comparable to the best fit value for the two-dimensional dissociation constant for EGFR-D107-fp (156 molecules/mm 2 ) ( Table 1), suggesting that concentration-dependent oligomerization might underlie ligand-independent EGFR phosphorylation.
To determine whether EGFR phosphorylation correlated with phosphorylation of downstream effectors in CHO cells, the influence of EGFR membrane concentration on two downstream signaling pathways (STAT and MAPK/Erk) was examined by Western blotting (Fig. 4). Increasing EGFR expression had no measurable effect on Erk1/2 phosphorylation or expression in the absence or presence of EGF (Fig. 4, A and D). Total STAT1 expression levels increased modestly with increasing EGFR expression, but, unlike EGFR, phosphorylation of STAT1 did not increase with increasing EGFR in the absence of ligand. The level of EGF-dependent pSTAT1 was enhanced in EGFR-YFP overexpressing cells, however (Fig. 4E). Thus, although higher EGFR concentrations result in higher basal levels of EGFR phosphorylation, this increase in ligand-independent phosphorylation does not correlate with any measurable increase in phosphorylation of downstream effectors.

Asymmetric kinase dimer interface mutations alter but do not destabilize ligand-independent EGFR oligomer ensembles
A key element of EGFR activation is formation of an asymmetric dimer interface between the N-lobe of one kinase domain and the C-lobe of another kinase domain. Single amino acid substitutions at these N-lobe (I706Q) and C-lobe (V948R) interfaces abolish EGF-dependent EGFR activity, but coexpression of these two point mutants restores activity because of their ability to complement each other's defect (37). To determine whether the asymmetric kinase dimer interface is involved in the formation of EGFR oligomers in the absence of ligand, the I706Q and V948R mutations were introduced into the EGFR-D107-fp variant and the oligomerization propensity of each variant measured by QI-FRET (Fig. 5). Neither variant had a strong effect on the propensity of EGFR to form ligandindependent oligomers. This is consistent with a previous observation that the EGFR-L704N variant, which disrupts the asymmetric kinase dimer interface and abolishes ligand-dependent EGFR phosphorylation, is 30-40% dimeric in the absence of ligand (54). These results contrast with previous observations that these mutations reduce ligand-dependent oligomers (41). In the absence of EGF, the dissociation constants did not significantly differ between the WT, I706Q, and V948R EGFR-D107-fp variants (Fig. 5, A and B and Table 1). Although these mutations did not alter EGFR oligomerization strength, they did alter the FRET efficiency of the oligomer ensemble. The FRET efficiency in the absence of ligand was much higher for both variants relative to WT EGFR-D107-fp, with a best-fit parameter for the intrinsic FRET of 0.46 and 0.70 for the I706Q and V948R variants, respectively, versus 0.35 for EGFR-D107-fp (Table 1). The FRET efficiency for both variants was also higher in the presence of ligand relative to WT EGFR-D107-fp in the presence of ligand (Fig. 5 and Table 1), indicating that disruption of the asymmetric kinase dimer alters the ensemble of intracellular domain interactions in ligand-driven EGFR oligomers.

Discussion
In this study we use QI-FRET to investigate homo-and hetero-interactions between EGFR and HER2 in vesicles derived from cell membranes. This approach allows measurement of FRET efficiency as a function of receptor concentration because of variable ERBB levels in vesicles derived from different cells. We observe concentration-dependent FRET efficiency values for several EGFR variants in the absence of ligand, The graph on the left shows the FRET efficiency ((FRET app 2 FRET prox )/X a )) (y axis) as a function of concentration (x axis). The data in B, D and F were projected onto the y axis to yield FRET histograms, plotted at the right. FRET app is the apparent FRET efficiency, FRET prox is the theoretical FRET efficiency that results from nonspecific interactions, and X a is the fraction of acceptor molecules in a given vesicle. Binned data points are shown as circles. Error bars represent the S.E. in x and y. The best fit to a monomerdimer equilibrium model is represented by solid lines. B, data for EGFR-D107-fp in the absence (black) and presence (green) of 100 nM EGF. Each dataset was derived from at least three independent biological replicates. The number of vesicles per experiment ranged from 25 to 100. which indicates dynamic equilibria between different EGFR association states and allows derivation of effective 2-dimensional dissociation constants. The homo-oligomerization strength of EGFR observed here, in both the presence and absence of ligand, is similar to previously published results (23,24,55). Mutations at the EGFR asymmetric kinase dimer interface alter the ensemble of ligand-independent EGFR oligomers, implying that ligand-independent EGFR oligomers are distinct from the active state dimer, consistent with previously published results (54). We also show that, although levels of phosphorylated EGFR increase with increasing EGFR concentrations, this phosphorylation does not correlate with any measurable increase in phosphorylation of either Erk1/2 or STAT1, indicating that receptor phosphorylation may be insufficient to trigger pathway activation as recently observed for artificially forced EGFR dimers (56,57).
Understanding the role of ligand-independent EGFR oligomers in regulating ERBB activity remains the topic of much investigation (18,19,41,54,(58)(59)(60). To estimate the oligomeric fraction of unliganded EGFR and HER2 in physiological membranes using the two-dimensional dissociation constants reported here, the concentration of ERBBs in cell membranes must be known. One recent study reported receptor concentrations in cell lines derived from human tumors as ;150 molecules/mm 2 (A549 cells) and ;650 molecules/mm 2 (A431 cells) (53). These cell lines have intermediate (A459, ;1 3 10 5 ) and high (A431, ;1 3 10 6 ) numbers of total receptors per cell. Uncertainty about the total surface area of a cell's plasma membrane precludes direct conversion between receptors per unit area and receptors per cell.
The presence of phosphorylated oligomeric EGFR in the absence of ligand (24,56,61) (Fig. 4) raises the question of whether EGFR is sampling the active dimer state, which involves a specific asymmetric dimer interaction between intracellular kinase domains (37). Amino acid substitutions at this interface that impair the ability of EGFR to become phosphorylated in response to ligand and preclude formation of the asymmetric dimer in crystals do not prohibit the formation of ligand-independent EGFR oligomers, however, implying the presence of an inactive EGFR oligomer that is distinct from the active dimer. This inactive oligomer appears to depend on the EGFR intracellular domain as deletion of intracellular domain results in loss or decrease of EGFR association as judged by FRET (62), single-molecule tracking (63), chemical crosslinking (64), and loss of negative cooperativity in a saturation binding experiment (18). The ERBB extracellular domains must also play a role in mediating or favoring this inactive oligomer; deletion of the extracellular domains leads to constitutive receptor phosphorylation (35), and mutations within the interface of the extended ECD dimer destabilize ligand-independent oligomers (23,63). No conserved interaction between the ERBB extracellular domains has been observed in crystals of tethered forms of ERBB ECDs, however, and it is not clear how they inhibit receptor activity in the absence of ligand.
In contrast, a number of interactions observed in crystal structures of inactive ERBB kinase forms have been suggested as reflective of inactive ERBB dimers (39,54,65,66). A consensus has yet to emerge on the physiological relevance of these dimers, however, and truncation of ERBB C-terminal tail domains in crystallized forms of inactive ERBB kinases may remove essential components of physiological interactions in the inactive state. Regions of the EGFR C-tail immediately following the kinase domain appear to play a role in stabilizing an inactive state as deletions or mutations in this region lead to enhanced or ligand-independent receptor activity (29,67,68). An attractive hypothesis is that the kinase distal region interacts with the kinase C-lobe in a fashion that competes with interactions made in the asymmetric kinase dimer interface (68). If so, this contact between the kinase and the tail might stabilize the inactive oligomer, which could underlie the observation that the kinase C-lobe mutation (V948R) alters the inactive oligomer ensemble as reflected in an increase in maximal FRET efficiency.
Conversion from inactive to active forms of ERBBs clearly involves coupled rearrangements and interactions between many structural elements. We provide evidence here that in the absence of ligand a substantial fraction of EGFR exists in an inactive oligomeric state that is distinct from the active state. We observe that HER2 is predominantly oligomeric in the plasma membrane, but our method does not resolve any structural or functional features of this oligomeric state. Future studies are needed to determine the structure and function of ligand-independent EGFR and HER2 oligomers and how they contribute to regulating receptor activity in normal and disease states. The graph on the left shows the FRET efficiency ((FRET app 2 FRET prox )/X a )) (y axis) as a function of concentration (x axis). The data were projected onto the y axis to yield FRET histograms, plotted at the right. FRET app is the apparent FRET efficiency, FRET prox is the theoretical FRET efficiency that results from stochastic, nonspecific interactions, and X a is the fraction of acceptor molecules in a given vesicle. Each dataset was derived from at least three independent biological replicates. The number of vesicles per experiment ranged from 25 to 100.

Plasmid construction
The coding sequences for variants of EGFR, HER2, EYFP, and mCherry were amplified using the PCR and cloned into pCDNA3.1(1). Fluorescent protein coding sequences were first inserted between the KpnI and XbaI sites. The coding sequences for EGFR and HER2 were then inserted between the NheI and KpnI sites. The dimerization-impaired EYFP-A206K variant was used. The missense variants EGFR-I607Q, EGFR-V948R, EGFR-L858R were generated using site-directed mutagenesis.

Cell culture and transfection
CHO cells and A431 cells were maintained in Dulbecco's modified Eagle Medium (DMEM F-12) supplemented with 2 mM glutamine, 5% FBS and were grown at 37°C, 5% CO 2 . Both cell lines were purchased from ATCC. For imaging experiments, CHO cells were seeded in 35-mm dishes at a density of 2 3 10 4 cells per well and grown for 24 h prior to transfection. Cells were transfected with plasmids using Lipofectamine 3000 (Life Technologies) according to the manufacturer's protocol.

Generation of monoclonal CHO cell lines expressing EGFR-EYFP
CHO cells stably expressing EGFR-EYFP were generated using the pCDNA3.1(1) manual as a guide. Briefly, cells transfected with EGFR-EYFP in the pCDNA3.1(1) vector were grown for several days in medium without antibiotics. G418 was added to a final concentration of 1 mg/ml and the cells were cultured until individual colonies appeared. All colonies were pooled into a single polyclonal culture, from which monoclonal populations were derived using FACS on a FACSCalibur instrument (BD Biosciences). Resulting cell lines were maintained in DMEM F-12 supplemented with 10% FBS, 1 mg/ml G418, and 25 mM HEPES, pH 7.2. Stable expression of EGFR-EYFP in these cell lines was assayed by confocal microscopy and Western blotting (see below).

Vesiculation, image acquisition, and analysis
All vesiculation procedures were performed as described in Refs. 48, 49, and 69. For ligand addition experiments, EGF was added to the vesicles to a final concentration of 100 nM and incubated for at least 1 h prior to imaging. Vesicle images were acquired using a Nikon C1 laser scanning confocal microscope at 603 magnification (water immersion objective). For each vesicle, three scans were recorded: a donor scan (l exc = 488 nm, l em = 500-530 nm), an acceptor scan (l exc = 543 nm, l em = 650 nm longpass), and a FRET scan (l exc = 488 nm, l em = 565-615 nm). Argon (488 nm) and He-Ne lasers were used as excitation sources. The image size was 512 3 512 pixels. The pixel dwell time was 1.68 ms, and the gains were set to 8. The Förster radius for the EYFP-mCherry pair was calculated to be 53.1 Å. All images were processed using a MATLAB program written in the Hristova laboratory which calculates the FRET efficiency All values in panels B-E represent integrated band intensities, and the highest value within the experiment for a particular axis was normalized to equal a value of 1. Green circles depict bands from wells that were treated with 100 nM EGF, white circles represent bands from wells which were left untreated. Data were fit to a straight line. Slopes that deviated significantly from zero (p , 0.002) are shown as dashed lines. B, phosphorylated-EGFR (pY-1068) is plotted on the y axis, total EGFR level on the x axis. C, phosphorylated-EGFR (pY-1068) divided by total EGFR level plotted on the y axis; total EGFR level on the x axis. D and E, phospho-Erk1/2 (D) and phospho-STAT1 (E), divided by the signal for total ERK1/2 and STAT1 proteins, respectively, as a function of total EGFR. Molecular mass markers (kDa) are indicated to the right of each Western blot. Western blots are representative of three independent biological experiments, all of which are plotted in panels B-E.
for each vesicle. A detailed description of the analysis is found elsewhere (46,47). Absolute protein concentration in the membrane was calculated by comparing the fluorescence intensity in vesicles with the intensities measured from a dilution series of fluorescent protein standards (EYFP and mCherry). Bleedthrough coefficients were calculated for each experiment (typically ;0.3 and ;0.2 for EYFP and mCherry, respectively). Where appropriate, the processed data (FRET versus concentration) were fit to a monomer-dimer equilibrium model using GraphPad Prism: The dimeric fraction for each vesicle, f D, is given by the equation where apparent E app is the apparent measured FRET efficiency, X a is the fraction of acceptor molecules in a given vesicle, and Ẽ is the FRET value within a dimer.

EGF stimulation of ERBB proteins in CHO cells
Recombinant human EGF was expressed in Escherichia coli and purified as described (70). Briefly, human EGF bearing an N-terminal hexa-histidine tag was purified using immobilized metal affinity chromatography. The hexa-histidine tag was removed by proteolysis with tobacco etch virus protease. Soluble, monomeric, human EGF was separated from tobacco etch virus protease by size-exclusion chromatography, concentrated, and flash frozen in liquid nitrogen. CHO cells were grown to 90% confluency, then transfected and grown for 14 h, then serum starved for 7 h at 37°C in Ham's F-12 supplemented with 1 mg/ml BSA. The cells were incubated in starvation media in the presence or absence of 100 nM EGF (5 min, 37°C). The medium was aspirated, the cells were washed twice with cold PBS (13, plus 1 mM Na 3 VO 4 ), then lysed for 15-20 min at room temperature in RIPA buffer (150 mM NaCl, 50 mM Tris, pH 8, 1% Nonidet P-40, 0.5% w/v sodium deoxycholate, 0.1% SDS, 1 mM Na 3 VO 4 ). The lysates were clarified by centrifugation and the total protein concentration of the supernatants was determined by BCA assay (Pierce, Life Technologies). Lysate protein concentrations were normalized, and the samples were analyzed by SDS-PAGE and Western blotting to detect phosphotyrosine (4G10, Millipore), EGFR (D38B1), EGFR pY1068, HER2 (29D8), or HER2-pY1221/1222 (all ERBB antibodies purchased from Cell Signaling Technology).

Data Availability
All data are included in this manuscript.  The graphs on the left show plots of the FRET efficiency (y axis, FRET app ) as a function of receptor concentration (x axis, units in molecules per mm 2 ) for I706Q (A) and V948R (B). Each protein is indicated above the graph. The black line shows the best fit line for EGFR-D107-FP from Fig. 2B. The dotted lines represent the best fit line for the I706Q (red) and V948R (blue) variants. As in Figs. 2 and 3, the FRET data in A and B were projected onto the y axis to yield FRET histograms, plotted at the right. Data were derived from three independent biological replicates, with 25-100 vesicles for each experiment.

Acknowledgments-Lily
necessarily represent the official views of the National Institutes of Health.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.