This work was supported by a grant from Sugen Inc. (to J. S.). The costs of publication of this article were defrayed in part by the payment of page charges.
Vol. 269, No. 13, Issue of April I, pp. 9752-9759, 1994 Printed in U.S.A.
THEJOURNAL OF BIOLMICAL CHEMISTRY Q 1994 by The American Society for Biochemistry and Molecular Biology, Inc.
Stabilization of an Active DimericForm of the Epidermal of an Inter-receptor Growth Factor Receptor by Introduction Disulfide Bond* (Received for publication, November 12, 1993)
Andrey SorokinS, Mark A. LemmonSO, Axel Ullrichn, and Joseph SchlessingerSll From the $Department of Pharmacology, New York University Medical Center, New York, New York 10016 and the TDepartment of Molecular Biology, Max-Planck-Znstitut fur Biochemie, A m Klopferspitz 18A, 8033 Martinsried, Germany
Populations of the epidermal growth factor receptor conclusions have subsequently been reached for the fibroblast (EGFR) with both high and low affinity for EGF are growthfactor and platelet-derivedgrowthfactorreceptors found on the surface of cells. It has been hypothesized (Amaya et al., 1991; Ueno et al., 1991). that the high-affinity state of the EGFR represents the Two properties of the EGFR which are essential for its norreceptor dimer and that this is also the kinase-active mal function have been ascribed to the dimeric form of the form. We describe herestudies aimed at addressing this receptor. These properties include activation of the tyrosine question directly.To stabilize homodimersof EGFR, we kinase moiety of the receptor, as well as the generation of have generated a mutated form of the receptor by in- high-affinity binding sites for EGF (Boni-Schnetzler and Pilch, serting a cysteine residuein the extracellular jux- 1987; Yarden and Schlessinger, 1987a, 198713). While the ditamembranous region, in order to cross-link the extra- meric form of the EGFR is generally considered to represent the active form of the receptor, no reliable comparison of the cellular domains of two receptors via disulfide bond formation. The mutation resultedin ligand-induced ap- functional activities of monomeric and dimeric EGFR has yet pearance of covalently linkedEGF receptor dimers and, been achieved. The main reason for this is the difficulty expein parallel, increasedthe number of high-affinity recep- rienced in attempting to separate the dimeric from the monotors present on the surface of cells expressing the mu- meric species of the receptor: previous studies have been pertated EGFR. Comparison of the tyrosine kinaseactivity formed under conditions which allowed interconversion of of thecovalentlylinkeddimericand the monomeric monomers and dimers. forms ofthe EGF receptor, separatedby sucrose density Scatchard analysis of EGF binding to cells has suggested gradient centrifugation, showed that the dimer sigwas that EGFR exists in two states on the cell surface; one with nificantly more active than monomer in the phospho- high affinity for EGF, and the other with lower affinity. The rylation of exogenous substrate. We conclude that the nature of the difference between these apparentpopulations is dimeric form of the EGF-receptor representsthe active not clear. It has been suggested that the dimeric form of the form, and that dimer formation is associated with the receptor represents thehigh-affinity state (Yarden and Schlessappearance of high-affinity binding EGF receptors on inger, 1987a, 1987b; Boni-Schnetzler and Pilch, 1987), and the the cell surface. monomeric form the low-affinity state. However, it has also been proposed that transmodulation (i.e. induction of disappearance) of the high-affinity cell surface EGF receptors by Ligand-induced dimerization is a property common to the platelet-derived growth factor and phorbol esters occurs withsignaling mechanisms of all known growth factor receptors out regard to the oligomeric state of the receptor (Northwood (Schlessinger and Ullrich, 1992). The first such receptor for and Davis, 1989; Walker and Burgess, 1991). which ligand-induced dimerization wasdemonstrated was that In an effort to overcome the difficulties experienced in atfor EGF.’ Although the precise regions of the receptor molecule tempting to characterize the dimeric form of the EGFR, we involved in stabilizing the dimer haveyet to be identified, have generated covalently linked dimers by employing the apEGFR dimerization is thoughtto be a critical earlyevent in the proach of disulfide cross-linking. This represents the least inresponse to stimulation with EGF.Accordingly, it has been vasive method via which receptors can be cross-linked coshown that coexpression of wild type EGFR with a mutated or valently. A similar approach has been employed for the neu truncated receptor lacking most of the cytoplasmic domain re- proto-oncogene. (Cao et al., 1992). In this case, dimerization sults in the suppression of signaling by the normal receptor: itself was not sufficient to activate thereceptor, but recovery of ligand bindinginduces the formation of an inactive het- dimerwasenhanced by the activating V t o E mutation at erodimer (Honegger et a l . , 1990; Kashles et a l . , 1991). Similar residue 664. The introduction of cysteine residues hasalso been used to probe the flexibility and ligand-induced structural * This work was supported by a grant from Sugen Inc. (to J. S.). The changes in the bacterial aspartate chemoreceptor (Falke and costs of publication of this article were defrayed in part by the payment Koshland, 1987;Lynch and Koshland, 1991; Pakula andSimon, of page charges. This article must therefore be hereby marked “aduer- 1992). tisement” in accordance with 18 U.S.C. Section 1734 solely to indicate In this paperwe report the isolation of stable EGFR dimers this fact. obtained by introducing a cysteine residue into the jwtamem5 Marion Abbe Postdoctoral Fellow of the Damon Runyon-Walter branous region of the extracellular domain between residues Winchell Foundation. 11 To whom correspondence should be addressed: Dept. of Pharmacol- 618 and 619 of EGFR. This insertion resulted in the covalent ogy, New York University Medical Center, 550 First Ave., New York, NY cross-linking of EGFR molecules. We found that the resulting 10016. Tel.: 212-263-7111; Fax: 212-263-7133. disulfide-mediated dimerization is ligand-dependent, and that, The abbreviations used are: EGFR, epidermal growth factor receptor; EGFR-IC, EGFR with introduced cysteine residue; PAGE, polyac- once formed, the disulfide-linked dimers were stable even after rylamide gel electrophoresis; PBS, phosphate-buffered saline solution. the removal of EGF by a low pH salt wash. This ability to
9752
Stabilization of an Active Form of EGF Receptor generate stable dimers, which do not have bound EGF, enabled us to measure the affinities for EGF of this otherwise inaccessible receptor state. Disulfide-mediated stabilization of EGFR dimers resulted in an increasethe inproportion of high affinity EGF-binding sites on the cell surface. Furthermore, separation of dimeric and monomeric EGFR by sucrose density gradient centrifugation revealed that the dimeric form of EGFR is that which is activeas a tyrosine kinase. These results represent the first direct confirmation that the dimeric form of the EGFR is responsible for high-affinity EGF binding, and is the alsoactive state of the receptor.
9753
cold PBS, and incubated in Dulbecco's modified Eagle's medium, containing 20 mM HEPES (pH 7.4) at 4 "C for an additional 30 min. By monitoring the binding of '""IEGF, we found that this combination of mild acidsalttreatmentwith Dulbecco's modified Eagle's medium washing was sufficient to remove more than 97% of the total EGF bound to the cells a t 4 "C without affecting cell permeability or EGF receptor binding (data not shown). '"I-EGF Binding Experiments-Human rEGF (Intergen) was iodinated with IODO-GEN (Sigma) to a specific activity of 80,000-120,000 cpdng. Parameters for EGF binding were determined by Scatchard analysis. Cells were plated in 24-well dishes and were allowed to grow t o confluence. Confluent cells were then incubated in Dulbecco's modified Eagle's medium containing 20 m,v HEPES and 1 mg/ml bovine serum albumin with '""IEGF (0.1-100 ng/ml) for 3 h at 4 "C. The level EXPERIMENTAL PROCEDURES of nonspecific binding was determined in every experiment by the addition of a 500-fold excess of unlabeled EGF together with '""IEGF to Construction of ExpressionPlasmids a n d Establishment of Cell the binding medium. Nonspecific binding was determined for every Lines-An EGF receptor mutant, with a cysteine residue introduced concentration of'"I-EGF used, and for every cell line tested. After into the juxtamembranous region of the extracellular domain, was conincubation, the cells were washed three times with ice-cold PBS, lysed structed by ligation of annealed syntheticoligonucleotides (encodingthe in 1 M NaOH, and radioactivity was counted in a y-counter. desired insertion) into a unique EcoRV restriction site. This site was Catalytic Oxidation of EGFR-IC Using an o-Phenanthroline Copper engineered into thecDNA for human EGFR by site-directed mutagenComplex-In an attempt to induce disulfide-linked dimerization in the esis (Amersham kit) using a synthetic oligonucleotide with the sequence 5'-GATGGACGGGATATCAGGCCCATTC-3'. The tetrapeptideCys-Alaabsence of EGF, cells were oxidized by ambient oxygen a t 22 "C with Ala-Ala was thus inserted between residues 618 and 619 of the mature o-phenanthroline-copper complex as a catalyst. The complex was prepared by mixing o-phenanthroline (from 1aM stock solution in ethanol) protein, immediately preceding the transmembrane domain. The syntheticoligonucleotides used for the insertion in these studies with a solution of 0.5 M CuSO, in a ratioof 1:2 (v:vl inHank's solution, 20 mM). had the sequences:5'-TGCGCCGCTGCA-3' and 5"TGCAGCCGCGCAand adjusting thepH to 7.96 with HEPES (final concentration Cells were washed with PBS and incubated with the o-phenanthroline3 ' . The desired alteration in the sequence of EGFR was confirmed by copper complex solution ( 3 mw) for 10 min a t 22 "C. '""IEGF binding DNA sequencing. The DNA encoding EGFR-IC was subcloned into a cytomegalovirus-based expressionvector for studies involving transient was then analyzed a s described above. Separation of EGF Receptor Dimers and Monomers by Sucrose Denexpression in 293 kidney cells (Margolis et al., 1992); it wassubcloned sity Gradient Centrifugation-200 pl of cell lysate was layered over a into a CVN vector (Riedel et al., 1986) for stable expression in transfected NIH 3T3 cells (which lack endogenous EGFR). Cells were trans4-ml sucrose density gradient ( 5 2 5 % )containing 20 mM HEPES (pH fected using the calcium phosphate precipitation as described (Chen 7.61, 0.1% Triton X-100, 150 mM NaCI, and 10% glycerol. Velocity sediand Okayama, 1987). Cells expressing EGFR were identified by immentation was carried out a t 4 "C for 22 h a t 190,000 x g in an SW60 rotor (Beckman). After development of the gradients, fractions were muno-blotting of cell lysates using antibodies raised against the carboxyl terminus of the receptor (Margolis et al., 1992). The cell lines 8 collected from the bottom of the tubes. Six initial fractions containing expressing EGFR-IC that were studied were EGFR-IC5 and EGFRdrops each were collected, followed by fractions of 16 drops each. ImIC17. The level of expression of the altered receptor in these cell lines munoblotting experiments or tyrosine kinase activity measurements was approximately 470,000receptors/cell in the caseof EGFR-IC5, and were performed directly upon the fractions. 120,000 receptors/cell in the case of EGFR-IC17. Measurement of Drosine Kinase Activity-Equal amounts of each Immunoblotting, Immunoprecipitation, and Peptide Mapping-Cell fraction from the sucrose gradient were added t o a solution containing lysis, immunoprecipitation, and immunoblotting procedures were per- 5 p~ ATP (with 1 pCi of ly-.'2P]ATPper sample), 1 mM MgCI,, or 5 mM formed as described (Margolis et al., 19891, except that p-nitrophenyl MnCI, and 300 p~ of K1 peptide (which represents the major autophosphosphate was omitted from the lysis buffer. Immunoblotting of phosphorylation siteofEGF receptor, at Y1173) (Honeggeret al., 1988). Each photyrosine-containing proteins wasperformed with a rabbitpolyclonal reaction was allowed to proceed for 1 min at 22 "C, then terminated by antibody (anti-P-Tyr); immunoblotting of EGF receptor was performed addition of gel sample buffer, and was then boiled for 5 min prior to with a rabbit polyclonal antibody raised against theCOOH terminus of analysis by SDS-PAGE. An SDS-PAGE gel consisting of a bottom onethe EGF receptor (anti-C); and immunoprecipitation of the EGF recep- third of 209' polyacrylamide overlayed by a 5-156 gradient was used to tor was performed with the monoclonal antibody 108 ( d b - 1 0 8 ) (Bellot separate the phosphorylated receptorfrom the phosphorylated peptide et al., 1990). Proteins were separated by electrophoresis on 4-109, grasubstrate and from free ATP and phosphate. dient SDS-PAGE gels. To estimate the ratio of dimeric to monomeric /'HIThymidine Incorporation-Cells were seeded at a density of EGFR, the hands corresponding to each species wereexcised from the 13,000 cells/well in 96-well dishes, grown overnight in 10%)calf serum gel, and the amount of radioactivityineachwasmeasuredwitha and then starved in0 . 4 8 calf serum. Eighteen hours after stimulation y-counter. Staphylococcus aureus V8 protease peptide mapping was with EGF or fetal calf serum, [,"Hlthymidine(0.5 pCi/ml, 6.7 Ci/mmol) performedwithimmunoprecipitatedEGFreceptormonomersand was added for another 16 h at 37 "C. The cells were then washed with dimers after invivo ['"Slmethionine labeling as described (Cleveland et PBS, trypsinized, and harvested using a PHD Cell Harvester (Camal., 1977). Phosphopeptide mapping was also performed by the same bridge Technologies Inc.). The amount of incorporated ["Hlthymidine approach, except thatmonomeric and dimeric EGFR were prepared by was quantitated by liquid scintillation counting (Pharmacia LKB). immunoprecipitation from unlabeled cells prior to autophosphorylation. In Vitro Autophosphorylation-For autophosphorylation assays, 10 RESULTS pl of HNTG (20 mM HEPES (pH7.51, 150 mM NaC1, 0.1% TritonX-100, Construction of a n EGF Receptor Able to Undergo Disulfide10% glycerol), containing 15 mM MnCI, and 2 pCi of Iy-"'P]ATP (6,000 CUmmol) was added to EGF receptor that had been immunoprecipilinked Dimerization-We have constructed an altered form of tated with monoclonal antibody mAb-108. The reaction mixture (in a the EGF receptor (EGFR-IC) containing a cysteine residue in final volume of 30 pl) was incubated for 15 min at 22 "C. The reaction the juxtamembrane region immediately amino-terminal to the was then stopped by the addition of 20 pl of SDS gel sample buffer. transmembrane domain (residue618). In order to study disulAnalysis by SDS-PAGE was performed both under reducing conditions fide-mediated dimerization of the altered receptor, EGFR-IC (+P-mercaptoethanol) and nonreducing conditions (-p-mercaptoethawas transiently expressed in 293 kidney cells, or stably exnol). Generation of Stable EGF-free Disulfide-linked EGFR-IC Dimerspressed in an NIH 3T3 (clone 2.2) cell line that lacks endogCells were plated in 24-well dishes and were allowed to grow to subenous EGF receptor. Upon transient overexpression of confluence. To induce the formation of disulfide-linked dimersof EGFREGFR-IC in 293 cells, stable dimers could be detected upon IC, cells were incubated with 400 ng/ml unlabeled EGF at 4 "C for 30 1). min. The cells were then washed with ice-cold phosphate-buffered sa- immunoprecipitationwithanti-receptorantibodies(Fig. This dimer formation was observed only when the receptor line solution (PBS) once, and incubated three times consecutively in contained a n introducedcysteineresidue;andaddition of ice-cold 0.2 .M sodium acetate buffer (pH 4.5) containing0.5 M NaCl ( 3 min each incubation), t o remove EGF. Cells were then washedonce with 1, l a n e P-mercaptoethanol was found to disrupt the dimer (Fig.
Stabilization of Active Form an
9754
EGFR-
EGFRW T C o nIt C rol
WT
IC
of EGF Receptor
EGFR-IC17
Control
EGFR-IC5 I
A. D -
WT 1-1
1
m-
D-
1
2
3
4
5
6
FIG.1. Expression of the wild type (WT) EGF receptor and the mutant EGFR-IC in 293 cells. EGFR (wild type) and EGFR-IC were immunoprecipitated from 293 cells with mAb-108, prior to autophosphorylation in the presence of Iy-:”PIATP. The reaction was stoppedby adding sample buffer with(lanes 1 3 )or without (lanes 4-6) a-mercaptoethanol. Monomers I M ) were separated from disulfide-linked dimers (Dl by SDS-PAGE in a 4-10% acrylamide gradient gel. The control lanes ( 3 and 6 1represent parallel immunoprecipitations from parental 293 cells.
2 ) . These observations indicate that dimerization is mediated
by disulfide bond formation between the introduced cysteine c residue of two adjacent receptors. The identity of the presumed dimer band as EGFR-IC dimer was confirmed by its detection EG F 0 10 5 0 100 0 10 5 0 100 0 10 50 100 with antibodies against EGF receptor, and by S. aureus V8 protease peptide mapping (data not shown). Frc. 2. Ligand-induced disulfide-mediated dimerization of wild type-EGFR Disulfide-linked Dimerization of EGFR-IC Is Ligan,d EGFR-IC.NIH 3T3 (clone 2.2) cells, stably expressing EGFR-IC were treated with the indicated concentrationsof EGF (in Dependent-Although significantlevels of disulfide-linked or ng/ml) for 20 mina t 4 “C, and then lysed. Cellular proteins were sepaEGFR-IC dimerscould be detected in immunoprecipitates from rated by nonreducingPAGEandweretransferredtonitrocellulose. 293 cells transiently overexpressing the altered receptor, con- Immunoblotting was performed with anti-EGFR antibody (anti-C) ( A ) stitutive dimerizationof EGFR-IC was not observed upon anal-or antiphosphotyrosine antibodies( B ) .The positions of monomeric ( M ) and disulfide-linked dimeric ( D ) EGFR are indicated by arrows. ysis of lysates from stably expressing cell lines. Under these conditions, the ability of EGFR-IC to form disulfide-linked homodimers was found to be clearly dependent upon stimulation sis, 10 mh1 iodoacetamide was included in all buffers (Fig. 4). with EGF (Fig. 2). The presence of iodoacetamide in the lysis buffer had no sigthe proportion of disulfide-linked Two possible explanations may be proposed for this observa-nificantinfluenceupon tion. EGF may inducea conformational change in the receptor, dimers that were detected. EGFR-IC Disulfide-linked Dimers Are Stable-Acid wash rewhich is required for efficient inter-receptor disulfidebond formationbetweenintroducedcysteineresidues.Alternatively, moval of EGF from the receptor has been reported to result in EGF may stimulate this disulfide bond formation indirectly, by the rapid monomerizationof receptor dimers (Sorkin and Carpromoting receptor dimerization. Thiswould have theeffect of penter, 1991).To determine whether this also occurs in the case increasing thelocal concentration of EGFR-IC surrounding the of disulfide-linked dimersof EGFR-IC, or whether the disulfide introduced cysteine, and may thus enhance disulfidebond for- bridge stabilizes the dimer, EGFR-IC5 cells were incubated mation (i.e. the reaction would be zero-order rather than sec- with EGF at 4 “C for 30 min. EGF was then washed away by ond-order in EGFR). In the case of transient expression of high treatment with an acetate buffer solution containing NaCI,at levels of EGFR-IC in 293 cells, the concentration of the receptor pH 4.5 (see “Experimental Procedures”). By contrast with the will be high. As a result, the amount of receptor dimer in the situation with wild type EGFR, even after removing EGF by absence of EGF will be increased by mass action, and thefor- acid/salt wash, stable dimers of EGFR-IC could still be detected mation of inter-receptor disulfide bondsis likely to be enhanced by anti-EGFR(anti-C)(Fig.5A)andanti-P-Tyr(datanot as described above. An analogous situation is suggested for shown) antibodies, although the proportion of dimer was retransformation resulting from overexpression of neu in NIH duced to some extent. Whereas the amount of 12”I-EGFbound to the receptor was reduced to 2-376 of its original value by this 3T3 cells (Lonardo et al., 1990). The effect of EGF upon the stimulation of disulfide-linked treatment, the abundanceof disulfide-linked EGFR-IC dimers EGFR-IC dimer formation was found to be both specific and fell only to 3 0 4 0 % of that detected before acid wash removal dose-dependent (Fig. 3). As expected, neither acidic fibroblast (Fig. 5, B and C ) . growth factor nor insulin were able to induce the formationof Disulfide-linked Dimerization Increases the Number of Highdisulfide-linked dimers of EGFR-IC (Fig. 4). To rule out the affinity EGF-binding Sites-Our initial experiments were depossibility that EGFR-IC dimerization occurred duringcell ly- signed to determine whether the introduction of a novel cys-
/
Stabilization of an Active Formof EGF Receptor
9755
4000 1
D+
E
3000
-
2000
-
0
1000
o
, 0
.
, 100
.
, 200
.
, 300
.
, 400
.
, 500
. 600
EGF nglml
3. Dose-dependenteffect of EGF on disulfide-linked dimerization. Cells were treated as described in thelegend to Fig. 2.4, and the amount of covalently linked dimer obtained upon treatment with different EGF concentrations was estimated by excising the bands representing EGFR-IC dimers from the a n t i 4 immunoblot, and counting their radioactive content ina y-counter. FIG.
1
2
3
4
5
6
Fw. 4. Specificity of EGF induced dimerization of EGFR-IC. An immunoblot of cell lysates treated as described in the legend to Fig. 2, blotted with anti-Cantibody. Lanes I and 3 show lysates of cells that have not been treated with EGF. Lanes 2 and 4 show lysates of cells treated with 100 ng/ml EGF. Lanes 3 and 4 are identical to lanes 1 and 2, respectively, except that 10 my iodoacetamide was included in all buffers: little difference in EGF-stimulated dimerization results. Lanes 5 and 6 show blots of lysates from cells treated with fibroblast growth factor (400 ng/ml) and insulin (1 pg/ml), respectively, in place of EGF.
teine residue, capableof mediating disulfide-mediated receptor dimerization, would alter the affinity of EGFR for EGF. Scat- phosphorylation states of monomeric and disulfide-linked dichard analysis of EGF binding to cells revealed only a small meric forms of EGFR were compared. In immunoblots of nondifference between those expressing wild type EGFR and those reduced lysates with anti-P-Tyr antibodies, i t was clear that expressing EGFR-IC (Table I). However, as discussed above, the dimeric formof the receptor hasa significantly higherlevel EGFR-IC does not dimerize when expressed in stable cell lines of autophosphorylation than does the monomeric form (Fig. 2). in the absence of EGF stimulation. We hypothesized that the One caveat in this approach for the comparison of the autoinduction of EGFR-IC dimerization would lead ato n increase in phosphorylation states of the different oligomeric states of the number of high-affinity EGF-binding sites. Therefore, we EGFR is that dimers whichare not disulfide-linked will dissoattempted to induce the formation of disulfide-linked dimers, in ciate to monomers under the conditions of nonreducing SDSthe absence of EGF, by the catalytic oxidation of sulfhydryl PAGE. As a result, the degree of autophosphorylation estigroups using an o-phenanthroline-coppercomplex as the cata- mated for monomeric EGFR will be affected by the comigration lyst (Kobashi, 1968) (see “Experimental Procedures”). This ap-of dissociated dimers. Therefore, to test further the hypothesis proach h a s previously been employed in the inductionof disul- that the EGFR dimer is the active form of the receptor, we fide bond formationincysteine-containingmutants of the separated dimeric from monomeric EGFR under nondissociatbacterial aspartate chemoreceptor (Falke and Koshland, 1987).ing conditions, using sucrose density gradient centrifugation. Unfortunately, in the present case, pretreatment of cells ex- As shown in Fig. 6, disulfide-linked dimers of EGFR-IC could pressing wild type EGFR or EGFR-IC with the o-phenanthro- be separated from monomers by velocity sedimentation on a line-copper complex resulted in a reduction in EGF binding, 5-25” (w:v) sucrose density gradient.To characterize the two perhaps as a result of interference with disulfide bonds that are isolated forms of EGFR, we measured their tyrosine kinase required for the structural integrity of the receptor. No en- activity using a synthetic peptide (K1) as a n exogenous subhancement of disulfide-linked dimers was detected (data not strate. The dimeric form of EGFR-IC was thus found to possess shown). much higher tyrosine kinase activity than themonomeric form As a consequence of these observations, another approach (Fig. 7A). was required for generating EGFR-IC dimers for EGF-binding One concern in these studies,as mentioned in the Introducstudies. To induce disulfide-mediated EGFR-IC dimerization, tion, is that exchange may occur between the monomeric and at 4 “C for 30 min. The cells were preincubated with EGF dimeric states of the receptor during the courseof the experibound EGF was then removed using an acidsalt wash,as de- ment. Since the dimeric form of the receptor studied here is scribed under “Experimental Procedures.” The appearance of covalently cross-linked, it cannot dissociate during the measurement. In addition, in order to prevent dimerization of the covalently linked EGFR-IC dimers was monitored, and EGFbinding studies were performed. As shown in Table I, the num- monomeric receptor during the tyrosine kinase measurements, ber of high-affinity receptors on the surfaceof cells expressing we reduced the final concentration of MgCl, to 1 mM, since high EGFR-IC that had been treated in this way increased approxiconcentrations of M e and MnZ+ havebeen shown to increase mately 10-fold, and in parallel with the occurrence of covalently the tyrosine kinase activity of EGF receptor,possibly by induclinked dimers of EGFR-IC. By contrast, cells expressing wild ing receptor aggregation (Mohammadi et al., 1993). Thus, the type EGF receptor showed only a reduction in the number of kinase activity measurements reported here were performed high-affinity binding sites in response to this treatment. Such under conditions which prevented exchange between the two a n EGF-induced reduction in the number of high-affinity EGF- states of the receptor. In the present study, the kinase activity binding sites has also been observed in other studies (Wiley et of monomeric receptor was found to be increased when assayed in the presenceof 5 mM MnCl, (data not shown). al., 1989; Walker and Burgess, 1991). Tyrosine Kinase Activity of Monomeric a n d Dimeric EGF As a n additional test of our hypothesis, we induced aggregaReceptors-In order to ascertain whether the dimeric form of tion of monomeric receptorsby treatment withmAb-108. It has EGFR represents the kinase-active form of the receptor, the been shown previously that bivalent antibodies are able to
Stabilization of a n Active Formof EGF Receptor
9756
A.
C.
B. In
40
D+
"-I
3
1
2
3
1
2
1
2
3
FIG.5. Stability of EGFR-IC disulfide-linkeddimers after removal ofEGF with an acidsalt wash. A, an immunoblot of cell lysates as presented in Fig. 2 A . Cella had been incubated with (lones 2 and 3 ) or without (lane 1 ) 200 ng/ml EGF for 1 h a t 4 "C. In the case of lone 3, the cells were subjected toacidsalt wash removal of EGF priorto cell lysis. I t is apparent that a significant amount of disulfide-linked dimer remains after this treatment. Panel C represents graphically the ratioof dimer tomonomer. Bands correspondingto dimer andmonomer were excised from the blot shown in A, and their radioactive content determinedby counting. The ratio of counts found in the dimer position to that found at the monomer position is plotted against lane number ( 1 , no EGF; 2, incubation with E G F 3 , incubation with EGF, followed by acidsalt wash). The proportion of EGFR-IC found a s disulfide-linked dimeris reduced by nearly 3-fold upon removal of EGF.Ponel B represents the amount of '"I-EGF bound to the receptor before and after the acidsalt wash treatment. TAIKEI Equilibrium binding parameters for wild type EGFreceptor and for EGFR-IC tnufant Parameters of "'I-labeled EGFbindingresults to wild typeand EGFR-IC expressed on transfected cells analyzed according tothe method of Scatchard (1949). K , and K2 are dissociation constants of high- and low-affinity binding sites, respectively; N,and N 2 are numbers of high- and low-affinity binding sites, respectively; EGFIASW means that the procedure of acidsalt wash removal of bound EGF following incubation with EGF was performed. The percentage quoted in the lastcolumn represents the percentage of receptors that bind EGF with high-affinity. Each value represents the mean result of a t least three independent binding experiments.
meric forms of the receptor, respectively (Schlessinger, 1988). Here, we report studies that strongly support this hypothesis, and present data suggesting that the dimeric receptor is the signaling active species. By introducing a novel cysteineresidueproximaltothe transmembrane domain of the receptor, we were able to generate stable, ligand-induced disulfide-linked dimers.Biochemical characterization of these stable receptor dimers enabled us to address directly the question as to whether dimerization of the receptor is responsible for both high-affinity ligand binding and activation of the tyrosine kinase catalytic domain. The ligand-induced disulfide-linked dimers were stable even after Cell line EGFIASW K, Nl KL N2 ?4 removal of EGF using an acid wash, enabling us to perform XI()""Y xlO'lcrll ~ 1 0 " Y' x1O'lrdl Scatchardanalysis of EGFbindingtocellsbearingstable HER3 1.0-5.0 5 2 4.0-6.0 2 115 2 10 4.3 EGFR dimers. These cells displayed a n increased number of HER3 + 1.0-4.0 20.4 5.0-10.0 80 f 10 0.5 EGFR-IC5 0.8-1.2 7 2 1 9.0-16.0 470 2 30 1.5 high-affinity EGF-binding sites. Furthermore, isolated stable EGFR-IC5 + 2.0-4.0 57 2 3 9.0-11.0 360 10 15.8 EGFR dimers showed tyrosine kinase activity that was markedly enhanced compared with that of a preparation of the activate EGFRby inducing receptor dimerization (Schreiber et monomeric receptor. Our finding that EGF binds with higher affinity to dimeric al., 1981, 1983; Spaargaren et al., 1991). Fig. 7B shows that than to monomeric EGFR in vivo is consistent with previous treatmentwithsuchanantibodyreducesthedifferencein measured tyrosine kinase activity between the preparations of studies of ligandbindingtodetergent-solubilizedreceptors dimeric and monomeric receptor, as a result of activation of the (Yarden and Schlessinger, 1987a, 1987b; Boni-Schnetzler and latter. This result further strengthens our conclusion that the Pilch, 1987). In addition, dimerization of the soluble extraceloligomeric state alone is responsible for the difference in kinase lular ligand-binding domain of EGFR has been shown to increase its affinityfor EGF (Zhouet al., 1993). However, several activity of the receptor preparations, and argues that the diof high-affinity meric form of EGFR is indeed the active form. As a final control, reports have suggested that the appearance as the state of cells expressing wild type EGFR were subjected to a n identical sites for EGF is dependent upon such variables sucrose density gradient analysis. The fractions corresponding receptor phosphorylation (Winston et al., 1991), association of to those that contained disulfide-linked EGFR-IC dimers were the receptor with the cytoskeleton (Van Bergen en Henegouwen al., 1989), as well as the state of receptor oligomerization analyzed, but no tyrosine kinase activity was detected (dataetnot a numshown). This result confirms that the enhanced phosphoryla- (Yarden and Schlessinger, 1987a, 1987b). Furthermore, cell-free membrane tion of theK1peptideseenin Fig. 7A is mediated by the ber of other studies in living cells and preparations have suggested that the appearance of high-affinEGFR-IC dimer. ity binding sites is independent of the oligomeric state of the DISCUSSION receptor (Northwood and Davis, 1989; Walker and Burgess, 1991). For example, platelet-derived growth factor stimulation The functional distinction between the high- and low-affinity of high-affinity EGF bindforms of the EGF receptor present on the surface of cells re- of cells resulted in the disappearance cells, but did not affect the proportion of EGFR found mains one of the major unresolved issues in considerations of ing to the the mechanismof signaling by the receptor.We have previously in a dimeric state,as assessed by cross-linking studies (Walker proposed that these forms represent the dimeric and mono- and Burgess, 1991). Similarly, phorbol ester treatment, which
Stabilization of an Active Form
A-
1
2
3
4
5
6
*.
M+
B.
9
8
7
of
IO 1 1
12
EGF Receptor 13
14
9757
15
16
17
18
19
b
-
?
4
1 3000
-
2000
-
1000
-
f
".
0
1
2
3
4
5
6
7
8
9
10
fraction
11
12 18 13 17161514
numbers
FIG.6. Disulfide-linkeddimers are separatedfrom monomersusingsucrose density gradient analysis. Cells were incubated with EGF Aliquots of collected fractions (see "Experimental a t 4 "C for 60 min and lysed prior to layeringover 4 ml of a 5-25% (w/v) sucrose density gradient. Procedures") were separated under nonreduced conditionsby SDS-PAGE on a 4-10% acrylamide gradient gel. Proteins were then transferredto nitrocellulose, and immunoblotted with anti-C antibodies (A). The amount of radioactivity in the bands corresponding to monomeric receptor (closed circles) and disulfide-linked dimers,(open circles) respectively, was estimated by excising the bands and counting in y-counter ( B ) .The positions of monomeric EGF receptor ( M ) and disulfide-linked dimers(D) are indicated by arrows. Only a portion of the sucrose density gradient is represented. Thefirst analyzed fraction (fraction 1 in the figure) corresponds to the seventh fraction collected.
reduces the number ofhigh-affinity sites for EGF,did not appear expressing wild type EGFR and EGFR-IC show very similar to reduce the proportion of EGFR found a s dimer (Northwood levels of EGF-induced DNA synthesis. Based upon these data, and Davis, 1989).However, it should be noted that those studies we reasoned that disulfide bridge formation resulted only in involved cross-linking of '"I-EGF to cell-surface receptorsusing the stabilization of EGF-induced dimers of the receptor, and disuccinimidyl suberimidate. Since the addition of "'1-EGF will had no deleterious effects upon its properties. alter themonomer/dimer equilibriumof the receptor (in favor of Our finding that the appearance of stable dimers coincided dimer), this approach, unlike theone used in our studies, ac- with an increase in the number of high-affinity binding sites on tually measures the ability of EGF to induce dimerization of the the surface of cells strongly supports the hypothesis that rereceptor rather than theproportion of dimeric receptor on the ceptor dimerization is required for high-affinity ligand binding. cell surface at the timeof the treatment. In thiscontext i t is of However, it remains a formal possibility that, in parallel with note that, although PMA treatment did not result in the apEGF-induced disulfide-mediateddimerization of EGFR-IC, parent disappearance of EGFR dimer in the reportmentioned other changes in receptor status occur, which could result in the above (Northwood and Davis, 1989), a marked phorbol 12-my- formation of the high-affinity sites.Walker and Burgess (1991), ristate 13-acetate-induced decrease was seen inI2'I-EGF label- for example, have proposed that high-affinity receptorsoccur in ing of dimeric receptor whenthe cross-linking experiment was the form of a complex with some other proteins that regulate performed at suboptimal (not saturating) levels of '"I-EGF. This high-affinity binding. Disulfide-linked dimer formation may also activate otherreresult is consistentwith our findings. Disulfide bonds are crucial for the folding, stability, and oli- ceptors. A point mutation (R129C) which introduces a cysteine gomerization of many membrane glycoproteins, and are usu- residue into the extracellularportion of the erythropoietin really formed in the oxidative environment of the lumen of the ceptor, for example, results in constitutive activation of this endoplasmic reticulum (Braakman et al., 1992). Disulfide-me- receptor (Watowich et al., 1992).This activation correlates with diated dimerization of EGFR-IC occurred on the cell surface disulfide-mediated homodimerization of erythropoietin receprather than the endoplasmic reticulum, and was induced by tor, indicating that,like EGFR, the dimeric form of the receptor EGF treatment. We were concerned initially that formation of is the signalingactive form. This approach of introducing cysthis disulfide bond would affect the folding of the receptor mol- teine residues into the ectodomain of cell surface receptors may ecule, and interferewith its correct function. A number of lines be of great value in the functional analysis of the consequences of evidence suggest that thisnot is the case. A comparison of the of receptor dimerization, a s well a s in the mapping of the intyrosine kinase activity in immunoprecipitates of wild type terface in dimeric proteins.The approach has been very informEGFR and EGFR-IC showed no difference. Furthermore, the ative, for example, in studying the intramembranous interacbiological function of the two receptors was comparable: cells tions between protomers of the bacterial aspartate receptor
Stabilization of an Active Form of EGF Receptor
9758
A 1 6 52 4 3
EGFR
190
-
125
-
88 6 5 56
79 8
10
13
12
11
14
15
16
18
17
19 2 0 21 2 2
. I . )
-
38 3 3.5
-
K1
B
1
2
3
4
5
6
7
8
9
10
11
13 12
14
15
16
17
I& 12229012
190
EGFR
12 5 88
65 56 38 3 3.5.
FIG.7. Comparison of t h e t y r o s i n ekinase activityof monomeric receptor withthat of the disulfide-linked dimer. Cells were treated a s described in the legend to Fig. 6, and the same fractionswcrc uscd for determination of tyrosine kinase activity. The ability of the molecules to phosphorylate endogenous substrate ( K 1 peptide) was tested in the presence of MgCl2 (1 m ~ n()A ) ,and after treatment with mAb-108 ( B ) .An SDS-PAGE gel consisting of a bottom one-third of 20% polyacrylamide overlayed by a 5-15% gradient was used to separate the phosphorylated receptor from the phosphorylated peptide substrate well as a s free ATP and phosphate. The positionsof the EGFR and the substrate peptide (K1) are marked by orrows.
(Pakula and Simon, 1992). All four subunits of the insulin receptor are linked into a hetero-tetrameric complex (a$,) by disulfide bonds (Pessin et al., 1985; Jacobs and Cuatrecasas, 1983). Thus, the insulin receptor may be consideredas a naturally occurring disulfidelinked dimer consistingof two @ subunit receptor halves. The association of these halves is required for the formation of a receptor with high affinity for ligand (Boni-Schnetzler et al., 1987). By contrast with the case of EGFR-IC, introductionof a novel cysteine residue into the juxtamembranous region of the neu proto-oncogene product (p185""'), another member of the EGF et receptor family, did not appear to activate the receptor (Cao
al., 1992). However, as in the caseof EGFR-IC, introduction of this cysteine did result in enhanced recovery of both ~185""' and its activated form ~185""' as disulfide-linked dimers (Cao et al., 1992). I t is well established that p185""" is more oligomerized than its untransforming counterpart (Weiner et al., 1989). That the transforming activityof the two altered forms of ~185""' was not affected by inducing disulfide-linked dimerization indicates that dimerization of this receptorper se is not sufficient for its activation. Rather, it might be suggested (Cao et al., 1992) that this reflects a requirement for a particular spatial relationship between the subunits of the receptor dimer. Alternatively, this result could be explained by abolition of the biological activity of the receptor upon formation of disulfide-
Stabilization of a n Active Form linked dimers. In the present study, it is clear that disulfidemediated dimerization of EGFR-IC does not abolish its tyrosine kinase activity. However, in this case, dimerization was ligand dependent, and thismay reflect the requirementfor a particular arrangement of two protomers with respect to one another for dimerization.This arrangementis presumablyinduced upon EGF stimulation. In conclusion, the studies reported here show that disulfide cross-linked dimers of EGFR-IC exhibited significantly higher levels of tyrosine kinase activity than themonomeric receptor, from which they had been separated. Our results demonstrate for the first time that the dimer is theactive form of the receptor under conditions in which exchange between the two oligomeric states was prevented. The ability of the isolated monomeric form of EGFR-IC to phosphorylate exogenous substrate was restored upon treatment with a bivalent antibody to the receptor. Thus, the kinase activity of EGF receptor monomers can be enhanced by inducing dimerization of the receptor in two independent ways. Thus, our results support the hypothesis that the dimeric form of EGFR represents the signaling active form of the receptor. Furthermore, we show that formation of the dimer is closely connected with the appearance of high-affinity EGF-binding sites. A full understanding of the structural basis for this dependence of kinase activity upon dimerization, andthe functional coupling betweenreceptor dimerization and its affinity for EGF is an important goal for future research. Acknowledgments-We thank Kiki Nelson for oligonucleotide synthesis, and Elena Sorokinafor help with sucrose gradients. W e also thank Eve Lowenstein and Ed Skolnikfor critical reading of the manuscript, and Moosa Mohammadi for valuable discussions. REFERENCES Amaya, E., Musci, T. J., and Kirschner, M. W. (19911 Cell 66, 257-270 Bellot, E , Moolenaar, W., Kris, R., Mirakhur, B., Verlaan, I., Ullrich A., Schlessinger, J., and Felder S. (19901 J. Cell Biol. 110, 491-502 Boni-Schnetzler, M., and Pilch, P.F. 11987) Proc. Natl. Acad. Sci. U. S. A. 84, 7832-7836 Boni-Schnetzler, M., Scott, W., Waugh, S.M., DiBella, E., and Pilch,P. F. 119871 J . Biol. Chem. 262,8395-8401 Braakman, I., Helenius, J., and Helenius, A. (19921 EMBO J . 11, 1717-1722
of EGF Receptor
9759
Cao, H., Bangalore, L., Dompe C., Bormann, B. J., and Stern D. F. (19921 J . Biol. Chem. 267,20489-20492 Chen, C., and Okayama, H. (19871 Mol. Cell. Biol. 7, 2745-2752 Cleveland, D. W., Fisher, S. G., Kirschner, M. W., and Laemmli,U. K. (1977)J. Biol. Chem. 252, 1102-1106 Honegger, A,, Dull, T. J., Szapary, D., Komoriya, A,, Kns, R., Ullrich, A,. and Schlessinger, J. (1988) EMBO J. 7, 3053-3060 Honegger, A. M., Schmidt, A., Ullrich, A,, and Schlessinger, J. (19901 Mol. Cell. Bid. 10, 4 0 3 5 4 0 4 4 Falke, J. J., and Koshland, D. E. (19871 Science 237, 1596-1600 Jacobs, S., and Cuatrecasas, P. (1983lAnnu. Rev. Pharmacol. Toxicol. 2 3 , 4 6 1 4 7 9 Kashles, O.,Yarden Y., Fisher, R., Ullrich, A,, and Schlessinger, J. 11991) Mol. Cell. B i d . 11,1454-1463 Kobashi, K. (1968) Biochim. Biophys. Acta 158,239-245 Lonardo, F., DiMarco, E., King, C. R., Pierce, J. H., Segatto, O.,Aaronson, S. A,. and DiFiore, P.P. (19901 New Biol. 2, 992-1003 Lynch, B. A,, and Koshland, D. E.(19911 Proc. Natl. Acad. Sci. U. S. A. 88, 1040210406 Margolis, B. L., Rhee, S. G., Felder, S., Lyall, R., Levitzki, A,, Ullrich, A., Zilherstein, A,, and Schlessinger, J. (1989) Cell 57, 1101-1107 Margolis, B., Hu, P., Katzav, S., Li, W., Oliver, J. M., Ullrich, A,, Weiss, A,, and Schlessinger, J. (1992) Nature 356, 71-74 Mohammadi, M., Honegger, A,, Sorokin, A,, Ullrich,A,, Schlessinger, J., and Hurwitz, D. R. (1993) Biochemistry 32,8742-8748 Northwood, I. C., and Davis, R. J. 11989) J. Bid. Chern. 264, 57465750 Pakula, A,, and Simon, M. (1992) Proc. Natl. Acad. Sci. U. S. A. 89, 4144-4148 Pessin, J. E., Mottola, C., Yu, K.-T., and Czech, M. P. (19851 in Molecular Basis of Insulin Action (Czech, M. P., edl pp. 3-29, Plenum Publishing Corp.,New York Riedel, H., Dull, T. J., Schlessinger, J., and Ullrich, A. (19861 Nature 324, 68-70 Scatchard, G. (1949)Ann. N. Y. Acad. Sci. 51,660-672 Schlessinger, J. (1988) Dends Biochem. Sci. 1 3 , 4 4 3 4 4 7 Schlessinger, J., and Ullrich, A. (1992) Neuron 9, 383-391 Schreiber, A. B., Lax, I., Yarden, Y., Eshhar, Z., and Schlessinger,J. (1981)J . Biol. Chem. 256, 716-720 Schreiber, A. B., Libermann, T. A., Lax, I., Yarden, Y., and Schlessinger,J. (19831J . Biol. Chem. 258, 846-853 Sorkin, A,, and Carpenter, G. (1991) J. Biol. Chem. 266, 23453-23460 Spaargaren, M., Defize, L. H. K., Boonstra, J.,and de Laat, S. W. 11991) J. B i d . Chem. 266, 1733-1739 Ueno, H., Colhert, H., Escohedo, J. A,, and Williams, L. T. (1991) Science 252, 844-847 Van Bergen en Henegouwen,P. M. P., de Kroon,J., van Damme, H., Verkleij, A. J.. and Boonstra, J. (1989) J. Cell. Biochem. 39, 455-465 Walker, F., and Burgess, A. W. (19911 J. Biol. Chem. 266, 2 7 4 6 2 7 5 2 Watowich, S. S., Yoshimura, A., Longmore, G. D., Hilton, D. J.,Yoshimura, Y., and Lodish, H. F. (1992) Proc. Natl. Acad. Sci. U . S. A . 89, 2140-2144 Weiner, D. B., Liu, J.,Cohen, J. A., Williams, W. V., and Greene,M. I. (1989)Nature 339,230-231 Wiley, H. S., Walsh, B. J., and Lund K. A. (1989) J. Biol. Chem. 264,18912-18920 Winston, J. T.,Olashaw, N. E., and Pledger,W. J. (1991)J. Cell. Biochern. 47,79-89 Yarden, Y., and Schlessinger, J. ( 1 9 8 7 4 Biochemistry 26, 1434-1442 Yarden, Y., and Schlessinger, J. 11987h) Biochemistry 26, 1443-1451 Zhou, M., Felder, S., Rubinstein, M., Hurwitz, D. R., Ullrich,A,,Lax, I., and Schlessinger, J. 11993) Biochemistry 32, 8193-8198
