CR1/CR2 Interactions Modulate the Functions of the Cell Surface Epidermal Growth Factor Receptor*

Recent crystallographic data on the isolated extracellular domain of the epidermal growth factor receptor (EGFR) have suggested a model for its activation by ligand. We have tested this model in the context of the full-length EGFR displayed at the cell surface, by introducing mutations in two regions (CR1 and CR2) of the extracellular domain thought to be critical for regulation of receptor activation. Mutations in the CR1 and CR2 domains have opposing effects on ligand binding affinity, receptor dimerization, tyrosine kinase activation, and signaling competence. Tyr246 is a critical residue in the CR1 loop, which is implicated in the positioning and stabilization of the receptor dimer interface after ligand binding; mutations of Tyr246 impair or abolish receptor function. Mutations in CR2, which weaken the interaction that restricts the receptor to the tethered (inactive) state, enhance responsiveness to EGF by increasing affinity for the ligand. However, weakening of the CR1/CR2 interaction does not result in spontaneous activation of the receptors' kinase. We have used an antibody (mAb 806), which recognizes a transition state of the EGF receptor between the negatively constrained, tethered state and the fully active back-to-back dimer conformation, to follow conformational changes in the wild-type and mutant EGF receptors after ligand binding. Our results suggest that EGFR on the cell surface can be untethered, but this form is inactive; thus, untethering of the receptor is not sufficient for activation, and ligand binding is essential for the correct positioning of the two receptor subunits to achieve kinase activation.

Over the last 20 years, the EGF 1 receptor has provided important opportunities for studying ligand activation of recep-tor-associated intracellular tyrosine kinases (1)(2)(3). Recently, the three-dimensional structures of the extracellular domains (ECDs) for several EGF receptor family members (EGFR, ErbB-2, and ErbB-3) have been reported (4 -9). These structures revealed two significantly different conformations for the EGF receptor ECD (4,5,9). In the crystal structure of the soluble, truncated ECD of the EGFR complexed with TGF-␣ (4) or with EGF (5), the ligand is sandwiched between the L1 and L2 (ligand binding) domains, the ECDs forming back-to-back dimers, primarily through the two interlocked CR1 (cysteinerich) domains; in contrast, in the crystal structure of the autoinhibited EGFR in complex with EGF, the ligand is bound only to the L1 domain, no dimer is present, and the main intramolecular interaction of the monomeric receptor occurs between the CR1 loop and the CR2 domain (9). In this structure, not only is the distance between L1 and L2 too great to allow simultaneous binding to one EGF molecule, but L2 is also rotated away from the L1-bound EGF. Thus, two critical features distinguish the autoinhibited (tethered) from the untethered form of the EGF receptor ECD: the absence of dimers and the inability to bind ligand with high affinity. Interestingly, the conformation of the truncated (8) and full-length (7) ErbB-2 ECD resembles the back-to-back EGFR dimer (4), whereas ErbB-3 ECD in the absence of ligand (6) has the same conformation as the tethered EGFR-ECD (9).
Work with the full-length, cellular EGFR has established a strong link between EGFR dimerization, high affinity binding, and receptor kinase activation. Whereas the crystal structures of the isolated ECDs provide an improved framework for the understanding of these observations (i.e. ligand will bind with higher affinity to the "untethered" form of the receptor, thus shifting the equilibrium away from the monomeric, autoinhibited receptor and favoring the formation of active dimers (9)), in the cellular environment the kinase and transmembrane domains of the EGFR also contribute to dimerization. Indeed, cell-surface dimers (or oligomers) can be detected in the absence of ligand, although the unligated dimers do not have tyrosine kinase activity (10 -16). Thus in the full-length, cell surface EGF receptor, ligand binding is required not only to drive dimerization but also for the formation of the kinase active conformation.
The structure and ligand binding properties of fragments or even full-length EGF receptor ECDs cannot unravel the complexity of signaling from cell surface-displayed receptors. In this report, in order to improve our understanding of the CR1-CR2 interactions on the processes that determine ligand binding, receptor conformational changes, receptor oligomerization, and the regulation of kinase activity, we have expressed fulllength EGF receptor mutants in mammalian (BaF/3) cells (17,18). BaF/3 cells express neither endogenous EGF receptors nor detectable levels of ligands that can perturb and/or activate recombinant (mutant) receptors. The availability of the CR1 loop and CR2 EGFR mutants and of the conformation-specific antibodies mAb 528 (19) and mAb 806 (20 -23) have allowed us to probe the determinants of tethering and to detect a major conformational transition when ligand binds to the receptor.

EXPERIMENTAL PROCEDURES
Reagents-Antibodies to the EGFR mAb 528 (19) and mAb 806 (20,21) were produced and purified in the Biological Production Facility at the Ludwig Institute for Cancer Research, Melbourne. The mAb 528 anti-EGFR hybridoma was purchased from ATCC. Anti-FLAG antibody M2 was purchased from Sigma; anti-phosphotyrosine (clone 4G10) and anti-EGFR (sheep polyclonal) were from Upstate Biotechnology, Inc. (Lake Placid, NY); and anti-phospho-p44/p42 MAPK antibodies and anti-MAPK antibodies were purchased from Cell Signaling (Beverly, MA). Horseradish peroxidase-coupled rabbit anti-mouse immunoglobulin and horseradish peroxidase-coupled rabbit anti-sheep immunoglobulin were obtained from Bio-Rad and Dako (Fort Collins, CO), respectively. Alexa 488-labeled anti-mouse immunoglobulin was purchased from Molecular Probes, Inc. (Eugene, OR). Phenylarsine oxide was purchased from Sigma. The water-soluble, homobifunctional crosslinking reagents BS 3 (spacer arm length 11.4 Å) and ethylene glycolbis(sulfosuccinimidyl succinate) (spacer arm length 16.1 Å) were obtained from Pierce.
Generation of EGFR Mutant Constructs-Single point mutations of the wild-type EGFR were generated using a site-directed mutagenesis kit (Stratagene, La Jolla, CA). The template for each mutagenesis was the human EGFR cDNA (accession number x00588) (24) as described in Ref. 4. Automated nucleotide sequencing of each construct was performed to confirm the integrity of each EGF receptor mutation.
Transfection of EGFR Constructs and Generation of Stable Cell Lines-Wild-type and mutant EGFRs constructs were transfected into the interleukin-3-dependent murine hemopoietic cell line BaF/3 as described in Ref. 17. Transfected cells were selected in G418 for 10 days. Viable cells were screened for EGFR expression by FACS analysis on a FACStar (Becton and Dickinson, Franklin Lakes, NJ) using antibodies to the FLAG tag (M2: 10 g/ml in PBS, 5% FCS, 5 mM EDTA) and/or to the EGFR extracellular domain (mAb 528: 10 g/ml in PBS, 5% FCS, 5 mM EDTA) followed by Alexa 488-labeled anti-mouse Ig (1:400 final dilution). Background fluorescence was determined by incubating the cells with an irrelevant, class-matched primary antibody. Positive pools were sorted for the appropriate level of EGFR expression on a FACS-DIVA (Becton and Dickinson). After final selection, mRNA was isolated from each cell line, and all mutations in the EGFR were confirmed by PCR analysis. All cells were routinely passaged in RPMI, 10% FCS, 10% WEHI3B conditioned medium (25) and 1.5 mg/ml G418.
Ligand Binding-Murine EGF, purified from mouse submaxillary glands (26), was iodinated using IODO-GEN (27) to a specific activity of 5-8 ϫ 10 5 cpm/pmol. Ligand binding to cells expressing the WT or mutant EGFR was determined at room temperature in the presence of the internalization inhibitor phenylarsine oxide (28) by cold saturation experiments as described in Ref. 17. All experimental points were prepared in triplicate. At the end of the incubation, the cells were pelleted and washed twice in ice-cold PBS before transferring to fresh tubes for counting in a Wallac WIZARD ␥-counter (PerkinElmer Life Sciences). Scatchard plots and estimates of ligand binding affinities and receptor numbers were obtained using the Radlig program (BioSoft, Cambridge, UK).
Receptor Cross-linking, Tyrosine Phosphorylation, and MAPK Activation-BaF/3 cells expressing the WT or mutated EGFR were incubated in medium without interleukin-3 and FCS for 3 h. Cells were collected by centrifugation, washed twice in PBS, and incubated in PBS at room temperature with or without EGF (100 ng/ml) for 10 min. In cross-linking experiments, the cells were incubated with 1.3 mM BS 3 or ethylene glycol-bis(sulfosuccinimidyl succinate) (Pierce) for 20 min at room temperature after PBS or EGF treatment.
Cells were lysed in SDS-PAGE sample buffer with or without reducing agent (100 mM ␤-mercapoethanol). Total cell lysates were analyzed directly by SDS-PAGE on 3-8% Tris acetate or 4 -12% bis-Tris gradient gels (Invitrogen) and transferred to PVDF membranes before immunodetection with anti-phosphotyrosine antibodies (4G10; Upstate Biotechnology; 1:1000 final dilution), anti-EGFR antibodies (sheep anti-EGFR; Upstate Biotechnology; 1:1000 final dilution) or anti-phospho-MAPK antibodies (1:1000 final dilution) followed by horseradish peroxidasecoupled anti-mouse, anti-sheep, or anti-rabbit Ig, respectively (all at 1:3000 final dilution). Reactive bands were visualized with ECL reagent (Amersham Biosciences). To determine specific tyrosine phosphorylation of the EGFR, membranes probed with anti-phosphotyrosine antibodies were stripped with a solution of 0.1 M glycine (pH 2.1) and reprobed with anti-EGFR or anti-phospho-MAPK antibodies. The films were scanned on an Amersham Biosciences scanning densitometer, and band quantitation was performed with ImageQuant using wide line peak integration.
Mitogenic Responses to EGF-Cells growing in log phase were harvested and washed three times to remove residual interleukin-3. Cells were resuspended in RPMI 1640 ϩ 10% FCS and seeded into 96-well plates using the Biomek 2000 (Beckman) at 2 ϫ 10 4 cells/200 l and incubated for 4 h at 37°C in 5% CO 2 . EGF was added to the first titration point and titrated in duplicate as 2-fold dilutions across the 96-well plate. [ 3 H]thymidine (0.5 Ci/well) was added, and the plates were incubated for 20 h at 37°C in 5% CO 2 before being harvested onto Unifilter 96 GF/c filter plates using an automatic harvester (all from Packard). The plates were air-dried for 1 h before the addition of scintillant (10 l/well). Incorporated [ 3 H]thymidine was determined using a ␤ counter (TopCount; Packard).
Reactivity with Conformation-specific Antibodies-Cells were preincubated with antibodies, EGF, or control medium prior to antibody staining and FACS analysis. Preincubation with antibodies (mAb 528, mAb 806, or a class-matched irrelevant antibody, all at 10 g/ml) was carried out at 37°C in RPMI, 10% FCS for times ranging from 30 min to 16 h. Preincubation with EGF (100 ng/ml in ice-cold FACS buffer) was carried out on ice for 20 min. After preincubation, cells were collected by centrifugation and stained with the control or test antibodies (all at 10 g/ml in FACS buffer for 20 min on ice, washed in FACS buffer) followed by Alexa 488-labeled anti-mouse Ig (1:400 final dilution, 20 min on ice). The cells were washed with ice-cold FACS buffer, collected by centrifugation, and analyzed on a FACScan; peak fluorescence channel and median fluorescence were determined for each sample using the statistical tool in CellQuest (Becton and Dickinson). Background (negative control) fluorescence was deducted from all measurements. The median fluorescence values were chosen as most representative of peak shape and fluorescence intensity and were used to derive the ratio of mAb 806 to mAb 528 binding.

RESULTS AND DISCUSSION
The aim of this work is to determine the role of CR1 loop/CR2 interactions on the conformational preferences, mechanism of activation, and signaling potential of the full-length, cell surface-expressed EGFR. We have introduced point mutations in the CR1 and CR2 domains, which would be expected to perturb the CR1/CR1 and/or CR1/CR2 interactions, and consequently alter the balance between the tethered, untethered, inactive, and/or active states of the EGFR. These constructs have been expressed in BaF/3, a hemopoietic cell line that, unlike the widely used 293 cells, does not secrete TGF-␣ or other EGFR ligands, is devoid of endogenous ErbB family members, and is therefore ideal for the biochemical characterization of the cell surface-associated EGFR (17,18). We have analyzed the effects of the mutations on the function of the EGFR by determining binding kinetics, dimerization, ligand-dependent tyrosine phosphorylation and signaling, and the ability to induce DNA synthesis in an EGF-dependent manner. These parameters are, however, indirect measures of receptor oligomerization, configuration, or conformational changes; therefore, we have also used the binding of two conformationally specific anti-EGFR antibodies, mAb 528 (19) and mAb 806 (20,23,29), as a tool to assess the effect of mutations on the "resting" conformation of the EGFR and on the dynamics of ligand-induced conformational changes.
Receptor Expression and Preliminary Characterization-Six point mutations have been analyzed in detail (see Fig. 1, A and B): three CR1 mutations at Tyr 246 (Pro, Trp, and Asp) and three CR2 substitutions at Asp 563 (to His), Glu 578 (to Cys), and Val 583 (to Asp). In an attempt to constrain the CR1/CR2 interaction with a disulfide link, we prepared a mutant with a substitution in each of CR1 and CR2 (Leu 245 to Cys and Glu 578 to Cys). After transfection in BaF/3 cells and selection in G418, receptor expression was monitored using the anti-FLAG anti-body M2 as well as the monoclonal antibody 528, which is directed to the extracellular domain of the EGFR, blocks ligand binding (19), and is reported to recognize only the native form of the receptor. Based on the reactivity with these antibodies, all mutant receptors appear to be correctly folded and are expressed at the cell surface. After multiple rounds of FACS sorting, we obtained cell lines expressing similar levels (20,000 -40,000 receptors/cell) of the mutant or WT EGFR (Fig.  2). It is essential that receptor expression is below 100,000 receptors/cell; transient expression experiments usually yield high levels of cell surface EGFR (Ͼ10 5 /cell); however, at these levels of expression there often is spontaneous activation (i.e. ligand-independent tyrosine phosphorylation) of the EGFR; thus, we have sought to avoid this complication by producing cell lines expressing Ͻ50,000 receptors/cell.
Ligand Binding by EGFR Mutants-From the crystal structures of the tethered (9) and untethered (4, 5) ECD of the EGFR, it has been postulated that the affinity of the ligand for the two forms will be quite different. Ferguson et al. (9) have reported that weakening the interaction between the CR1 and CR2 loops increases the apparent affinity of the EGFR-ECD for EGF; however, the link between tethering of the CR1 loop and the CR2 domain and ligand binding affinity is based on data obtained by BIAcore analysis of the isolated EGFR-ECD (9,30). Kinetic binding data for full-length EGFR at the cell surface yield affinity constants that are at least 2 orders of magnitude lower (20 pM to 2 nM) compared with 20 -350 nM for the EGFR-ECD. The binding kinetics of EGFR to its ligands in a cellular context are complicated by structure-independent factors such as local receptor density, oligomerization state, and interactions with cytosolic or cytoskeletal elements (31)(32)(33)(34) as well as by modifications in the kinase, transmembrane, and/or C-terminal domains (35)(36)(37)(38)(39). Therefore, it is important to measure the effects of CR1 and CR2 mutations on the ligand binding affinity of the receptor in intact cells. To prevent internalization while assessing ligand binding at a physiological temperature, affinity determinations were carried out in the presence of 30 M phenylarsine oxide (28); under these assay conditions, internalization of the EGFR was reduced to Ͻ1% (data not shown) The results of Scatchard analyses of EGF binding to WT and mutant EGFR are presented in Table I and Fig. 3 and  are summarized below. CR2 Mutations-The V583D and D563H mutations were designed to disrupt the CR2/CR1 loop interactions (see Fig.  1B). In the tethered conformation, the ␥-methyl groups of the Val 583 side chain are in close van der Waals contact with Tyr 246 ; substitution with the Asp ␥-carboxyl should disrupt the CR1/CR2 interface. Similarly, the ␥-carboxyl of Asp 563 is hydrogen-bonded to Tyr 246 in the tethered conformation, and substitution of the aspartate side chain with the imidazole of His will weaken the interaction. In cells expressing V583D, there is a significant increase in the proportion of high affinity EGF binding sites compared with cells expressing the WT EGFR (12.6 versus 2.6%, respectively) as well as an apparent increase in the affinity (3 versus 30 pM). This trend is also observed in the D563H mutant, although in this case the difference from WT was not statistically significant (Table I). An increase in the proportion of high affinity sites is an indication of a shift in the equilibrium toward the untethered states of the receptors, supporting the assumption that V583D and D563H mutations weaken the CR1/CR2 interactions.
In order to investigate the possibility of creating a disulfide bond to covalently link CR1 and CR2, we identified two residues that have appropriate distance and side chain orientation in the tethered conformation: Leu 245 and Glu 578 . Initially, we made the single mutation E578C and then the double mutation L245C/E578C. The Glu 578 side chain is close to the side chains of both Leu 245 and Phe 248 , so the E578C substitution might be expected to improve the packing of the CR1 loop/CR2 interface by increasing hydrophobic interactions with these residues. Experimentally, the E578C mutation completely abolishes high affinity EGF binding without affecting the number of low affinity sites (Table I). The introduction of a cysteine in this position does not appear to affect the folding of either or both cysteine-rich domains; the conformation-dependent antibody mAb 528 binds to the mutant receptor, and its phosphorylation and signaling are still dependent on EGF (see below). We believe that the effect of this mutation is limited to a strengthening of the CR1 loop/CR2 domain interaction.
The L245C/E578C double mutant, designed to lock the receptor in the inactive "tethered" state via a disulfide bond, also bound EGF with low affinity. However, the B max determined from Scatchard plots is ϳ10-fold lower than the number of receptors estimated by surface antibody staining or immunoblotting, suggesting that the majority of the receptor population fails to bind EGF with any significant affinity; thus, this mutant may be partially misfolded. Attempts to activate this receptor with reducing agents have not been successful. Until we are confident that L245C/E578C EGFR is folded correctly, results obtained on this mutant need to be interpreted with caution.
CR1 Loop Mutations-We introduced three different amino acid substitutions, Phe, Trp, and Asp, for Tyr 246 (see Fig. 1). The crystal structures suggest that Tyr 246 is critical for both the CR1/CR1 and CR1/CR2 interactions (Fig. 1B). In the tethered configuration, the CR1 loop interacts closely with the CR2 domain; Tyr 246 hydrogen bonds with the carboxyl side chain of Asp 563 , which in turn is held in place by a salt bridge with the ⑀-amino group of Lys 585 . Mutation of Tyr 246 to Phe removes the hydrogen bond, so the tether will be weaker. The Trp 246 mutant is too large to fit into the CR2 binding site, and indeed it would disrupt the Lys 585 -Asp 563 salt bridge. Replacing Tyr 246 with Asp will lead to the loss of hydrophobic packing as well as to a strong repulsion with Asp 563 . Thus, all mutations should render the tethered conformation less favorable. To activate the EGFR kinase, the back-to-back dimer must form, so that the phenolic oxygen of Tyr 246 hydrogen bonds to the opposing chain (Fig. 1B). Indeed, in the presence of ligand, the phenolic oxygen is hydrogen-bonded to the backbone at residues Gly 264 and Cys 283 . These two hydrogen bonds will be missing in the Phe 246 and Asp 246 mutants. The packing between Tyr 246 and the opposing chain is tight, with no room for a Trp residue; it is expected that the Trp 246 dimer would not be closely packed. Experimentally, all three mutations resulted in loss of high affinity EGF binding (Table I and Fig. 3), suggesting a severe impairment of the CR1/CR1 interaction, which is not compensated by the untethering of CR1/CR2 binding.
Taken together, these observations confirm that ligand binding to the "tethered" form of the EGFR occurs with low affinity, probably due to the failure of EGF to bind to both L-domains simultaneously. Inspection of the crystal structure indicates that the ligand binding surfaces of both domains are available in both the tethered and untethered conformations, so low affinity binding presumably reflects binding to either domain or to both domains independently. Clearly, untethering can increase the proportion of receptors available for high affinity binding. It is interesting to note that the high affinity conformation requires the CR1 loop, suggesting that dimerization is a necessary part of the high affinity state. Surprisingly, in a report recently published, Mattoon et al. (40) describe very different effects of mutations in the CR2 domain on ligand binding; substitutions designed to untether the intramolecular interactions appear to lower the affinity of the mutant receptors for EGF, although the affinity constants for these mutants are not reported. The ligand binding experiments described in Ref. 40 were, however, performed under conditions permissive of internalization, resulting in Scatchard plots that are nonlinear and difficult to interpret. Furthermore, this apparent loss of high affinity binding is not reflected in the dose responses for productive EGFR signaling as measured by kinase activation and MAPK phosphorylation (40).
Receptor Dimerization-The ligand-induced CR1/CR1 interaction is necessary for the formation of an active EGFR complex; deletion of the CR1 loop abolishes the ability of the EGFR-ECD to dimerize, even in the context of the full-length EGFR (4). Clearly, mutations in the CR1 and CR2 loops have significant effects on EGF binding affinity ( Fig. 3 and Table I); we were interested to determine the effect of these mutations on basal and ligand-mediated dimerization and kinase activation. Cells were treated with EGF and the homobifunctional, cellimpermeable cross-linker BS 3 for 30 min at room temperature. Cell lysates were separated by SDS-PAGE and immunoblotted with either anti-EGFR or anti-phosphotyrosine antibodies. The results are shown in Fig. 4 and summarized below.
CR1 loop mutants had reduced ligand dependent dimerization; in particular, the Y246D mutation completely abolished ligand-dependent dimerization. However, basal dimerization was only marginally affected; this points to a different role of Tyr 246 in the spontaneous and ligand-mediated dimerization interface. Given the complete lack of detectable dimers in the ⌬-CR1 loop receptor, in which the whole of the CR1 loop is deleted (4), it is possible that other regions in this loop contribute to the formation of the unligated dimer. The phosphotyrosine content of both the monomeric and dimeric Tyr 246 mutant receptors was also reduced, suggesting that, even when  . 3. Scatchard analysis of EGF binding to WT and mutant receptors. Ligand binding affinities were determined at a fixed concentration of 125 I-EGF by competition with unlabeled EGF (see "Experimental Procedures"). The plots were generated from the raw data using the "Kell for Windows" version of the RadLig program (BioSoft). dimers do form, the ECD conformation does not permit kinase activation; although some spontaneous dimers could be detected in the Y246W mutant, in the absence of ligand there is virtually no phosphorylation of the dimer. Clearly, the formation of ECD-cross-linkable dimer is reduced in all of the Tyr 246 mutants. It is interesting to note that the phosphotyrosine content of the Tyr 246 mutant monomers after EGF stimulation is particularly affected (Fig. 4C); since the monomers presumably are generated from dimers that have failed to cross-link, they may reflect a subpopulation of molecules with altered (weaker) interactions in the dimeric complex. Whether the Tyr 246 mutations overall affect the stability of the dimer or prevent a reorientation of the dimer subunits or the formation of higher order oligomers necessary for kinase activation cannot be addressed directly in our experimental system. CR2 mutants had normal levels of basal and ligand-dependent dimerization. We did not detect significant increases in the proportion of dimers for mutant EGFR in which the CR1/CR2 tether had been weakened, suggesting that, even when the mutations lead to untethering, the formation of the BS 3 -crosslinkable dimeric complex is dependent on the binding of ligand. The mutation of Glu 578 to Cys introduces an unpaired cysteine and could conceivably lead to the formation of a disulfidebonded dimer. We have investigated this possibility using cross-linkers of different spacer arm length (BS 3 , 11.3 Å; ethylene glycol-bis(sulfosuccinimidyl succinate), 16.1 Å) as well as analyzing noncross-linked dimers under reducing and native conditions (data not shown). We found no evidence of spontaneous dimerization of the E578C mutant and conclude that Cys 578 does not lead to the formation of interchain disulfide bonds.
Ligand-dependent Tyrosine Phosphorylation and MAPK Signaling-CR1 loop/CR2 interactions appear to stabilize a kinase-inactive conformation of the EGFR and prevent spontaneous activation (9). We monitored basal and EGF-dependent tyrosine phosphorylation as well as MAPK activation in cells expressing the mutant receptors. The results are presented in Fig. 5. Ligand binding causes some increase in the phosphotyrosine content of most mutant receptor molecules; however, the specific activation of individual mutants (measured by the ratio of tyrosine phosphorylation to receptor protein and by specific activation of MAPK) (Fig. 5, B and C) varied significantly. All CR2 mutants are activated by ligand at levels similar to the WT receptor. Even E578C, which has only low affinity sites and hence should occur predominantly in the tethered (inactive) form, can be fully stimulated at high concentrations of EGF (16 nM). We tested the correlation between ligand binding affinity and signaling of the CR2 mutant receptors by exposing the cells to increasing concentrations of EGF (30 pM to 100 nM) and monitoring the induction of tyrosine phosphorylation and MAPK activation (Fig. 6). In E578C-EGFR-expressing cells, peak phosphorylation of the EGFR and the signal transducers Shc and MAPK was only achieved at significantly higher concentrations of EGF compared with the WT. In contrast, activation of the receptor for the V583D and D563H-EGFR-expressing cells occurred at lower concentrations of EGF then WT EGFR (Fig. 6, B and C). These results support the concept that mutations in the CR2 domain affect binding affinity but not the subsequent events that trigger receptor function. Even at saturating amounts of EGF, all Tyr 246 mutants have severely reduced receptor tyrosine phosphorylation and MAPK activation (Fig. 5); the ability to form a productive CR1/CR1 loop interaction is critical for kinase activation. Other point mutations in the CR1 loop or its docking regions (Y251A and F263A) appear to have minimal effects on EGFR signaling (5). However, when both the CR1 loop and its docking site are disrupted (e.g. Y251A/R285S double mutant) (5), signaling is disrupted completely. These authors also reported a reduction in the level of ligand binding, suggesting that engagement of the dimerization docking site may influence the ability of L1 and L2 domains to reorient in response to EGF.
Mitogenic Signaling from EGFR Mutants-We have tested FIG. 4. Dimerization of WT and mutant EGFRs and specific phosphotyrosine content of receptor complexes. Quiescent cells were treated with EGF (100 ng/ml, 16 nM) or control buffer. The homobifunctional, cell-impermeable cross-linker BS 3 was added immediately, and the incubation continued for 30 min at room temperature. After quenching the reaction, the cells were lysed, and cellular proteins were separated by SDS-PAGE and transferred to PVDF membrane for immunoblotting. A, immunodetection of EGFR protein (top) and phosphotyrosine (bottom). The PVDF membrane was stripped after exposure to the anti-phosphotyrosine antibody and reprobed with the anti-EGFR antibody. B, ratios of dimer to total EGFR (dimer plus monomer) with and without EGF stimulation, determined by quantitative scanning densitometry as described under "Experimental Procedures." C, ratios of phosphotyrosine content to EGFR monomer and dimer protein, determined by quantitative scanning densitometry as above.
the EGFR mutants for their ability to induce de novo DNA synthesis following exposure to increasing concentrations of EGF, using a [ 3 H]thymidine incorporation assay. The results are presented in Fig. 7 and Table II. First, none of the cell lines exhibited ligand-independent [ 3 H]thymidine incorporation; it is clear that even when the tether between the CR1 loop and CR2 has been weakened, mitogenic signaling requires EGF binding for the activation of the receptor. Although generally the EC 50 values for EGF correlate well with the high affinity receptor occupancy (cf. Table I and Table II), in the case of E578C there is a 10-fold difference between the concentration of EGF required for half-maximal [ 3 H]thymidine incorporation and for half-maximal occupancy of the receptors. We have established that, in the BaF/3 cell lines expressing WT EGFR, as few as 500 receptors/cell need to be activated to achieve a half-maximal response to EGF (see Ref. 17 for methodology); this threshold is reached for WT EGFR at ϳ15 pM and for the E578C cells at ϳ80 pM EGF (Table II). Using the same calculations (based on the total number of receptors/cell and the fractional occupancy at each EGF concentration), we have estimated that the Y246W mutant should reach half-maximal response at an EGF concentration of ϳ60 pM and that the FIG. 5. Ligand-dependent tyrosine phosphorylation and MAPK activation. A, quiescent cells were exposed to EGF (100 ng/ml) for 10 min at room temperature and then lysed directly in SDS-PAGE sample buffer. Proteins were separated on 4 -12% gradient gels, transferred to PVDF membranes, and probed with antibodies to phosphotyrosine (top) or to phospho-MAPK (bottom). The blots were stripped and reprobed with anti-EGFR antibodies or anti-MAPK antibodies, respectively (not shown), to allow the determination of specific protein phosphorylation as described under "Experimental Procedures." B, ratios of phosphotyrosine to EGFR protein for WT and mutant receptors. C, ratio of phospho-MAPK to total MAPK protein.
FIG. 6. Dose response of EGFR activation in CR2 mutants. Cells expressing the WT or CR2 mutant receptors were rendered quiescent by growth factor and serum withdrawal and then exposed to control buffer or to increasing concentrations of EGF (0.03-100 nM). A, total cell lysates were analyzed by SDS-PAGE on 4 -12% gradient gels, followed by immunoblotting with anti-phosphotyrosine, anti-EGFR, or anti-phospho-MAPK antibodies. B and C, the films were scanned for densitometric quantitation of the reactive bands, and the phospho-Shc and phospho-MAPK data were plotted as percentage of maximal band intensity against EGF concentration. Closed circles, WT EGFR; dark triangles, D563H-EGFR; light triangles, Y246D mutant should reach half-maximal response at an EGF concentration of ϳ400 pM. The complete lack of response to EGF of these mutants in the mitogenic assay reflects an inability of these EGFRs to form a productive signaling unit rather than a simple loss of ligand binding affinity.
Antibody Monitoring of EGFR Conformations-The results presented so far support a model where interactions between the CR1 loop and the CR2 domain constrain the EGFR to a low affinity, kinase-inactive state (9), and the interaction of the CR1 loops within the back-to-back dimer is necessary for ligand-induced kinase activation of the EGFR (4,5). It is still unclear whether an "intermediate" state also exists and what its properties may be. The concept of an intermediate state is relevant to the biology of EGFR. The WT EGFR on the cell surface appears to be kinase-inactive, whether dimeric or monomeric; clearly, if there is a dynamic equilibrium between the tethered and untethered states, this transition does not lead directly to kinase activation. Our data suggest that transition from tethered monomer to untethered monomer (or dimer) in the absence of ligand involves the formation of a higher affinity state. Concentrations of ligand in vivo are so low that occupancy of the low affinity state is likely to be minimal. However, if the intermediate affinity state is present, ligand occupancy would be expected to drive the receptor complex to the kinaseactive form. None of the experiments we have described above, using EGFR mutants, addresses directly the existence of such an intermediate state of the EGFR. We therefore approached this problem by using antibodies specific to two different domains of the EGFR.
Monoclonal antibody 528 (19) has been used as a competitive antibody for EGF binding to the human EGF receptor. mAb 528 reacts with the ⌬2-7 EGFR (which lacks most of the L1 and CR1 domains (41)) and interferes with ligand binding to the  a The response of WT EGFR-BaF/3 cells at 1 nm EGF was taken as maximal.
b EC 50 was determined from the dose-response curves as shown in Fig 7. c The concentration of EGF needed to occupy 500 receptors/cell was calculated from the K d and B max data obtained by Scatchard analysis, using a plot of receptor occupancy versus EGF concentration. The number of EGFR occupied at each concentration of EGF was calculated from the formula, (͓L͔/͓L͔ ϩ K d1 ) ϫ R 1 ϩ (͓L͔/͓L͔ ϩ K d2 ) ϫ R 2 , where ͓L͔ ϭ EGF concentration; K d1 and K d2 are equilibrium binding constants; and R 1 and R 2 are the number of high and low affinity receptors. The data are representative of at least three separate experiments.
WT EGFR, suggesting that the epitope resides on the L2 domain. Reactivity with mAb 528 of all of the mutants used in this study is unimpaired, and receptor numbers determined by FACS analysis using mAb 528 or Scatchard analysis using 125 I-EGF are generally in agreement.
Monoclonal antibody 806 recognizes the ⌬2-7 truncated EGFR as well as a subpopulation of WT EGFR in cells overexpressing the receptor (23,29). mAb 806 is active as an antitumor agent in glioblastoma xenografts expressing ⌬2-7 EGFR or carcinomas, which overexpress the WT EGFR (22,42,43). It was postulated that this antibody selectively recognizes an activated form of the receptor (44). The recently identified peptide sequence that forms the mAb 806 epitope 2 is shown in magenta on the crystal structures of the EGFR-ECD (Fig. 1B). In contrast to the lack of mAb 806 reactivity in many epithelial cell lines expressing low EGFR numbers, BaF/3 cells, which express ϳ40,000 EGFR/cell, have weak but detectable mAb 806 binding. BaF/3 cells expressing a similar number of the ⌬2-7 receptors (which lack most of the L1 and CR1 domains) bind mAb 806 strongly. Intriguingly, the ⌬CR1 loop mutant (which lacks amino acids 244 -259 (4)) also has strong mAb 806 reactivity (Fig. 8).
Assuming that mAb 528 can recognize all of the correctly folded EGFR on the cell surface, using FACS analysis it is possible to determine the proportion of EGFRs reactive with mAb 806 by calculating the ratio in median fluorescence of mAb 806 to mAb 528; a direct comparison is possible because we use both antibodies at saturating concentrations, binding is detected by the same secondary antibody, and FACS detection is linear in the range used. Using this analysis, the proportion of receptors reactive with mAb 806 varies from 6 -8% for the WT EGFR to 70 -90% for the ⌬2-7 and ⌬CR1 loop EGFR (ratios of mAb 806 to mAb 528 binding of 0.06 -0.08 and 0.069 -0.98) ( Fig. 8 and data not shown). These data suggest that the epitope for mAb 806 may be masked by the native conformation of the WT receptor but exposed by deletion of the CR1 loop. Positioning of the epitope on the crystal structures of the EGFR-ECD (Fig. 1B) shows that it is likely to be buried at the CR1/CR1 interface in the back-to-back dimer form and at least partially buried in the tethered form of the receptor. Only in the putative "intermediate" untethered form of the receptor, where it is not masked by CR1 loop/CR2 or CR1/CR1 loop interactions, is the mAb 806 epitope likely to be available. This antibody therefore could provide a sensitive conformational probe for analyzing tethered, untethered, and fully active EGFR complexes.
To test this hypothesis, we have monitored the reactivity of mAb 806 with cells expressing WT EGFR before and after preincubation with mAb 806 or with EGF. WT EGFR/BaF cells were exposed to mAb 806 in the presence of the internalization inhibitor phenylarsine oxide (28) at 37°C (to maximize the energy of the system) or to EGF at 4°C to allow formation of the kinase-active state but completely exclude internalization. mAb 806 treatment did not alter the total number of EGFR (as determined by 125 I-EGF binding), and under both conditions more than 95% of the EGFRs were present at the cell surface (data not shown). After pretreatment with mAb 806, EGF, or control buffer, the WT EGFR reactivity with mAb 528, mAb 806, or control antibodies was measured by FACS analysis. Table III shows the changes in median fluorescence channel caused by the pretreatment with mAb 806 or EGF as well as the ratios between mAb 806 and 528 reactivity. This method of presenting the data was chosen to overcome variations between experiments in absolute median fluorescence values (which are very sensitive to small changes in the laser current and in the detector settings) and to allow pooling of the experimental data. Preincubation of the cells at 37°C for 1 h with 10 g/ml mAb 806 more than doubled the reactivity with mAb 806 without affecting 528 reactivity; thus, the ratio between the two antibodies was significantly elevated. Preincubation with mAb 528 under identical conditions had no effect on subsequent mAb 528 or mAb 806 binding (data not shown). In separate experiments, we proved that the enhanced mAb 806 binding was not attributable to lack of saturation, since increasing the concentration of mAb 806 (from 10 to 50 g/ml) or the time of exposure (from 20 min to 1 h) during the second incubation had negligible effects (data not shown). The effect of preincubation  III Variation in median fluorescence channel for WT EGFR-BaF/3 cells upon preincubation with mAb 806 or EGF BaF/3 cells expressing the WT EGFR were preincubated with control buffer, with mAb 806 (10 g/ml at 37°C for 1h (a)), or with EGF (10 ng/ml at 4°C for 15 min (b)). Cells were then probed with either mAb 806 or mAb 528 (both at 10 g/ml) followed by Alexa488-labeled anti-mouse Ig as described under ''Experimental Procedures.'' Cells were analyzed on a FACScan, and median fluorescence values were obtained using the statistical analysis program in CellQuest. Median fluorescence values after mock preincubation were 112 Ϯ 21 for mAb 528, 7 Ϯ 1.9 for mAb 806, and 0.5 Ϯ 0.3 for the control (class-matched) irrelevant antibody. Negative control values were subtracted from all data. The results are presented as positive or negative percentage changes in median fluorescence for the test samples compared with the mock samples. The ratios between median fluorescence for mAb 806 and for mAb 528 are also presented. Conversely, preincubation of the cells with EGF drastically decreases the reactivity with mAb 806. Internalization of the receptor under these conditions is Ͻ5% and hence cannot contribute significantly to the decrease in mAb 806 binding. In these experiments, the reactivity with mAb 528 also was reduced by ϳ20% after binding of EGF either through steric hindrance or masking of the epitope. Taken together, these results point to selective recognition by mAb 806 of an untethered, unligated form of the receptor Analysis of mAb 806 binding to the CR1 loop or CR2 mutants (Table IV) supports this hypothesis; mAb 806 reactivity was at least double that of WT EGFR in mutants with weakened CR1 loop/CR2 interaction (V583D and D563H) and around 3-fold higher than WT EGFR for receptors incapable of forming the CR1/CR1 interaction (Tyr 246 mutants). Incubation with EGF had opposite effects on the two classes of mutants; EGF reduced the reactivity with mAb 806 of the receptors capable of forming the active dimer (WT and all of the CR2 mutants), whereas the reactivity with mAb 806 was unchanged or even enhanced for the CR1 loop mutants. The effect of EGF on these mutants is consistent with an EGF-mediated untethering of a weak CR1 loop/CR2 loop interaction, accompanied by a failure to form the CR1/CR1 loops interaction. Modulation of mAb 806 reactivity by EGF correlates well with the ability or the failure of the mutant EGFRs to activate the EGFR kinase, as determined by tyrosine phosphorylation and by DNA incorporation (see Fig. 7 and Table II).
Our data are consistent with a model in which mAb 806 recognizes preferentially an untethered form of the EGFR, which is yet to be configured to the back-to-back dimer conformation. Thus, mAb 806 can be used as a tool to monitor conformational changes within the receptor upon ligand binding. The transient, untethered and unligated conformations of the EGFR would represent, at any one time, a small proportion of the total EGFRs but would be present in detectable amounts on cells overexpressing the receptor, as reported in the literature (22,23,29). Our data may also help explain the ability of mAb 806 to suppress tumor formation; in cells expressing the ⌬2-7 EGFR, binding of the antibody would sterically hinder formation of the kinase-active conformation of the receptor complex, whereas in cells overexpressing the WT EGFR it may trap the untethered EGFR form and prevent interaction between the CR1 loops and consequent activation. This hypothesis is consistent with the reported decrease in kinase activation of the ⌬2-7 EGFR after treatment with mAb 806 (43).
Role of CR1 and CR2 in EGFR Signaling-Mutations designed to test the role of the intrareceptor and interreceptor tethers (4,6,9) in the context of the full-length, cellular EGFR indicate that the proportion of EGFR in the high affinity binding state is strongly affected by the CR1 loop/CR2 tether, presumably reflecting the relative positioning of the L1 and L2 domains. Weakening of the CR1/CR2 tether increases the pro- FIG. 9. EGFR conformations and activation. The EGFR undergoes a major conformational change during the transition from the low affinity to the high affinity state. The low affinity conformation (A) is tethered by intramolecular interactions between the two cysteine-rich domains CR1 and CR2. The tethered monomer (A) is in equilibrium with either the tethered dimer (B) or a high affinity untethered monomer (F). It appears that transmembrane (T m ) and/or kinase domains drive the formation of both the tethered dimer (B) and the untethered dimer (C). The tethered dimer (B) is depicted in the schematic diagram with intermolecular contacts between both the ECD and kinase domains; however, further experiments are needed to substantiate the relevance of this model. The tethered forms of the receptor are low affinity. The untethered monomer and dimer have higher affinity, but it is not clear whether monomer and dimer have the same affinity. The intracellular kinase domains of the untethered dimer are not activated until ligand (e.g. EGF or TGF-␣) binding induces a further reorientation in the dimer-ligand complex (D). The receptor-ligand complex is capable of forming higher order oligomers (e.g. tetramers) (E); it is possible that higher order oligomers are required for full kinase activation. The ligand binding affinity is further modulated by inside-out signaling (e.g. ATP), and it is possible that the unligated high affinity dimer (C) or the untethered monomer (F) is the target of intracellular affinity modulators. Although ligand binding and dimerization/ oligomerization lead to kinase activation and substrate phosphorylation, signaling from the receptor is also regulated by internalization, degradation, and dephosphorylation. portion of high affinity sites and strengthening the CR1 loop/ CR2 tether abolishes high affinity binding (Table I). Notwithstanding the significant differences in ligand binding affinities between the full-length cellular receptor and the isolated ECD, the CR1 and CR2 interactions drive the same relative changes in the two molecules (cf. Ref. 9 and our data). Modulation of EGFR affinity by intracellular components (34,36,45,46), which have been attributed to modification of the juxtamembrane or kinase domains of the EGFR, must then reflect an altered balance between tethered and untethered states.
Ligand-independent dimerization (or oligomerization) of the EGFR is not significantly affected by mutations in the CR1 loop or CR2 domains. Weakening of the CR1 loop/CR2 tether does not lead to constitutive dimerization; nor does strengthening of the tether decrease it (Fig. 4). Thus, even when the CR1 loop is available for interreceptor interactions (as suggested by the mAb 806 results in Table IV), productive dimerization and activation (assessed indirectly by phosphotyrosine content) do not occur without ligand binding. However, ligand-mediated EGFR dimerization and activation are affected by the mutation in the CR1 loop. Thus, the constitutive and ligand-induced dimers are not equivalent, and ligand binding is strictly required for the fine positioning of the receptor subunits and consequent kinase activation (16). Tyr 246 in the CR1 loop appears crucial for the formation of the activated complex; our results obtained using the conformation-specific mAb 806 (Table IV) point to an inability of Tyr 246 mutants to orient the dimeric complex correctly.
We were able to monitor significant changes in the conformation of the EGFR using an antibody, mAb 806, which appears to recognize selectively the untethered but inactive form of the EGFR. Disruption of the CR1/CR2 interactions increases mAb 806 reactivity, whereas ligand binding decreases it (Table  IV). Furthermore, the receptors with the mutations of Tyr 246 most likely to disrupt the CR1/CR1 interaction (to tryptophan and aspartic acid) show a significant increase in mAb 806 reactivity after EGF binding, confirming that the activated CR1/CR1 orientation is compromised.
Whenever the ability to form an active dimer is maintained, the responses to EGF are dictated solely by the balance between affinity and receptor number. We have shown that, in BaF/3 cells expressing ligand-activable EGFRs, as few as 500 receptors/cell need to be occupied to stimulate half-maximal DNA synthesis (Table II), and this correlates with threshold stimulation of downstream signaling effectors such as Shc and MAPK (Fig. 5). Thus, whereas EGFR phosphorylation itself continues to increase with receptor occupancy, the signaling pathways are fully activated at a much lower ligand concentration; indeed, mitogenic stimulation occurs at concentrations of EGF where phosphorylation of Shc and MAPK, but not EGFR phosphorylation, are easily detectable.
In contrast to the report by Matton et al. (40), we observe a direct correlation between mutant receptor affinity, biological responses (signal activation and mitogenicity) and antibody reactivity; the agreement between all parameters of EGFR behavior analyzed so far strengthens our conclusion that the unligated, untethered receptor is primed to respond to EGF, whereas the tethered receptor is in a dormant state that requires high levels of ligand to initiate signaling.
It is becoming clear that the EGFR can exist in multiple states, each with different ligand binding characteristics and potential for activation by ligand; minor shifts in the equilibria between these forms can have significant repercussions for EGFR biology, particularly considering how few receptors need to be activated to fully trigger the downstream signaling cascades. We have attempted to summarize our understanding of EGFR alternative conformations and their role in receptor activation in Fig. 9. Some of the alternative configurations depicted in Fig. 9 are still only speculative but are consistent with our understanding of the system. Much still remains to be experimentally determined before we can claim to understand the processes that induce EGFR kinase activity. The FRET and FLIM methodologies (47,48) will be of aid in determining the cell surface oligomerization state of the EGFR in the untethered, tethered, unligated, and ligated forms and whether unligated dimers (or higher order oligomers) do indeed constitute a significant proportion of the EGFR population. The availability of the three-dimensional structures for two different conformational states of the EGFR-ECD has given us a revolutionary insight into parts of the process required to activate this receptor. However, the final triggering of kinase activity is still hidden behind the oligomerization and/or reorientation of the ligated dimer. It will be exciting to discover exactly how ligand binding initiates kinase activation and signaling by the EGFR.