The enhanced tumorigenic activity of a mutant epidermal growth factor receptor common in human cancers is mediated by threshold levels of constitutive tyrosine phosphorylation and unattenuated signaling.

Deregulation of signaling by the epidermal growth factor receptor (EGFR) is common in human malignancy progression. One mutant EGFR (variously named ΔEGFR, de2-7 EGFR, or EGFRvIII), which occurs frequently in human cancers, lacks a portion of the extracellular ligand-binding domain due to genomic deletions that eliminate exons 2 to 7 and confers a dramatic enhancement of brain tumor cell tumorigenicity in vivo. In order to dissect the molecular mechanisms of this activity, we analyzed location, autophosphorylation, and attenuation of the mutant receptors. The mutant receptors were expressed on the cell surface and constitutively autophosphorylated at a significantly decreased level compared with wild-type EGFR activated by ligand treatment. Unlike wild-type EGFR, the constitutively active ΔEGFR were not down-regulated, suggesting that the altered conformation of the mutant did not result in exposure of receptor sequence motifs required for endocytosis and lysosomal sorting. Mutational analysis showed that the enhanced tumorigenicity was dependent on intrinsic tyrosine kinase activity and was mediated through the carboxyl terminus. In contrast with wild-type receptor, mutation of any major tyrosine autophosphorylation site abolished these activities suggesting that the biological functions of ΔEGFR are due to low constitutive activation with mitogenic effects amplified by failure to attenuate signaling by receptor down-regulation.

Ligand binding to wild-type epidermal growth factor receptor (wt EGFR) 1 results in receptor dimerization, kinase activation, and autophosphorylation that provides both docking sites for proteins involved in signal transduction and exposure of endocytic and lysosomal targeting sequence codes required for receptor internalization and down-regulation (1). The biochemical and biological roles of each autophosphorylation site in wt EGFR have been explored by mutational analysis, and mutation of any single autophosphorylation site does not significantly abrogate binding of the activated receptor to specific SH2-containing proteins associated with distinct signaling pathways (2). Likewise, such single mutations are generally incapable of reducing the biological functions of the receptor in in vitro models (3,4). Correspondingly, the mitogenic and transforming activities of wild-type receptor were diminished only when combinations of favorable autophosphorylation sites (i.e. Tyr-1068, Tyr-1148, and Tyr-1173) were mutated (3), suggesting that the autophosphorylation sites of wt EGFR may have less specificity for signaling proteins and can compensate for each other. Sites of tyrosine phosphorylation may be provided via heterodimerization with other members of the erb B family of receptors (5)(6)(7). Point mutations that inactivated the tyrosine kinase activity of wt EGFR eliminated occupancyinduced receptor internalization (8,9,40,41), whereas mutant receptors lacking multiple autophosphorylation sites also lacked the ability to undergo ligand-induced endocytosis, suggesting that kinase-regulated receptor internalization is mediated by the phosphotyrosine residues (1,10). The role of the kinase and phosphotyrosine residues in EGFR down-regulation may be more indirect. Evidence suggests that autophosphorylation of EGFR results in a more open conformation (11) to expose the otherwise cryptic internalization/lysosomal targeting motifs to the receptor trafficking machinery, thereby triggering receptor down-regulation processes. Receptors lacking internalization sequences can lead to cell transformation (12). Thus, the normally strict controls on cell proliferation may loosen when regulatory constraints intrinsic to the EGFR structure are relaxed by these mutations.
Mutations of the EGFR gene occur in many types of human tumors where growth deregulation is pervasive (13)(14)(15). In glial tumors of the central nervous system, abnormalities of the EGFR are restricted to grade III (anaplastic astrocytoma) and especially grade IV (glioblastoma multiforme) disease (16 -18). Several clinical and histopathological studies have shown that such abnormalities are related to a shorter interval to relapse and decreased survival rates, suggesting that they play an important role in glioma malignancy progression (18,19). Nearly 50% of grade IV tumors have significantly amplified EGFR genes, and of these, the majority also show rearrange-ments of the gene so that these gliomas express both wt EGFR and the mutant forms (16,20,21). The most common of the rearrangements are genomic alterations that eliminate a DNA fragment containing exons 2-7 of the gene and generate a mutant receptor (variously called ⌬EGFR, de2-7 EGFR, or EGFRvIII) with an in-frame deletion of 267 amino acids from the extracellular domain (21)(22)(23)(24). This specific genetic alteration is also found frequently in lung and breast cancers (15,25). In previous studies we found that expression of ⌬EGFR in glioblastoma cells markedly enhanced tumor cell growth in nude mice (26), but the molecular mechanisms of the enhanced tumorigenicity have not been elucidated. Two defects have been described for the ⌬EGFR, lack of EGF binding and constitutive autophosphorylation that was unaffected by ligand but whose extent was low compared with that of ligand-activated wt EGFR in the same cell. The importance of autophosphorylation as a marker of constitutive kinase activation of ⌬EGFR to tumor growth was thus unclear.
Here we report that ⌬EGFR are expressed on the cell surface but are defective in down-regulation due to low rates of receptor endocytosis. Both the intrinsic tyrosine kinase activity and each of the major sites of tyrosine autophosphorylation located in the regulatory carboxyl terminus are essential for the enhanced tumorigenesis characteristic of ⌬EGFR. Unlike wt EGFR, a single point mutation in any of the major autophosphorylation sites is sufficient to abrogate the in vivo growth advantage conferred by ⌬EGFR. Thus deletion of exons 2-7 abolished ligand binding but created a constitutively active EGFR that was strictly dependent for biological activity on intact autophosphorylation. Defective attenuation due to failure to down-regulate activated receptors appears to be the major mechanism that enhances tumorigenicity of these weakly phosphorylated EGFR mutants. Thus deletion of exons 2-7 results in a partially active EGFR that couples to mitogenic signal transduction but not to vesicular trafficking pathways.

EXPERIMENTAL PROCEDURES
Materials-Anti-EGFR antibodies included mouse monoclonals 528 and 225 (27) and 13A9 (28) directed against the ectodomain, rabbit polyclonal sera, Ksm, directed against residues 642-663 in the submembrane domain (provided by J. Kyte, University of California at San Diego, La Jolla, Ca), and mouse monoclonal C13 directed against residues 991-1018 in the carboxyl terminus. The monoclonal antibodies specific for ⌬EGFR, D806, D1133, and D124, were generated against the fusion junction in the ectodomain of the mutant receptor. 2 The monoclonal antiphosphotyrosine antibody PY20 was from Transduction Laboratories (Lexington, KY), and the rabbit polyclonal antiphosphotyrosine sera were prepared as described (29).
FACS (Fluorescence-activated Cell Sorter) Sorting and Analysis-The human glioblastoma cell line U87MG (ATCC) was infected with mutant EGFR viruses and selected in a medium containing 400 g of G418 per ml (26). 5 ϫ 10 5 cells in 100 l of staining buffer (phosphatebuffer saline (PBS), 1% calf serum, 0.1% sodium azide) were exposed (30 -60 min) to anti-EGFR monoclonal antibody 528 which recognized both wild-type and mutant EGFR, or cells were exposed to monoclonal antibody D806 specific for mutant receptors, and then exposed to fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin antibody (PharMingen, Inc.) for 30 -60 min. Stained cells were analyzed with a FACSort (Becton Dickinson). For cell sorting, sodium azide was omitted from the staining buffer, and the brightest 5% of the stained cells were selected with a FACStar (Becton Dickinson). The sorted cells were cultured and reanalyzed. The number of DK and DY mutant receptors on the cell surface was estimated by comparing the fluorescence intensity of these cells with that of U87MG.⌬EGFR cells which expressed 2 ϫ 10 6 ⌬EGFR per cell (26).
Immunoprecipitation and Western Blotting-Cells (4 ϫ 10 5 cells per well in a 12-well plate or 2.5 ϫ 10 6 cells per 100-mm dish) were cultured overnight in medium containing 10% dialyzed fetal bovine serum. The cells were exposed to 100 ng/ml EGF (Collaborative Biomedical Products) for 5 min at 37°C. For Western blotting, cells were washed with PBS, lysed in a PBS buffer containing 0.06 M Tris-Cl, pH 6.8, 1.3% SDS, 12.8% (v/v) glycerol, 1.3% 2-mercaptoethanol, 2 mM Na 3 VO 4 , 5 g/ml leupeptin, 5 g/ml antipain, 5 g/ml pepstatin, and 0.25% bromophenol blue, and then boiled. The lysates were electrophoresed through SDS, 7.5% polyacrylamide gels and the proteins transferred to nitrocellulose membranes (Bio-Rad). For immunoprecipitation/Western blotting, cells were lysed in a PBS buffer containing 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 0.2% sodium azide, 0.04% sodium fluoride, and 5 g/ml each of the protease inhibitors. Cellular debris was removed by centrifugation at 12,000 ϫ g for 5 min at 4°C. EGFR were immunoisolated using a ⌬EGFR-specific monoclonal antibody, D1133, or anti-EGFR monoclonal antibody, 528, electrophoresed, and transferred to nitrocellulose membranes. Blots were probed with horseradish peroxidase-conjugated anti-phosphotyrosine monoclonal antibody PY20, and the proteins were visualized using the enhanced chemiluminescence detection system (Amersham Corp.). The blots were stripped and reprobed with the C13 anti-EGFR monoclonal antibody in the enhanced chemiluminescence system. EGFR concentration and phosphorylation were determined by scanning the autoradiograms with a densitometer. Multiple exposures were obtained to ensure linearity of response, and relative phosphorylation was calculated under these conditions. The level of EGFR phosphorylation was normalized to the protein concentration. The enhanced chemiluminescence was employed to minimize the possibilities of residual signal when stripping blots probed with 125 I-protein A.
For additional quantitation of phosphotyrosine content, confluent monolayers of U87MG or U87MG.⌬EGFR cells on 100-mm dishes were untreated or treated with EGF or 225 IgG (27), rinsed with PBS, and extracted with 500 l of boiling SDS sample buffer (5% SDS, 50 mM Tris-Cl, pH 6.8, 5% 2-mercaptoethanol, 1 mM Na 3 VO 4 ). Equal amounts of protein were loaded on 5-15% gradient gels, transferred to nitrocellulose in the presence of 0.1 mM Na 3 VO 4 , and blocked with 2.5% BSA, 0.1 mM Na 3 VO 4 , and 0.005% Tween 20. Tyrosine phosphate was detected using 5 g/ml affinity purified antiphosphotyrosine polyclonal 2 H.-J. S. Huang, E. Stockert, and L. J. Old, unpublished data. antibodies in conjunction with 50 ng/ml 125 I-Protein A for 30 min. Dried blots were exposed to storage phosphor plates followed by imaging using a Bio-Rad G250 Molecular Imager. Profiles from the appropriate lanes were used to determine the difference between the phosphotyrosine pattern of either control or U87MG cells expressing ⌬EGFR using the Molecular Analyst software package.
Fluorescence Microscopy-Cells were plated on fibronectin-coated coverslips 48 h prior to the experiment and were either untreated or treated with 100 ng/ml EGF for 2 h and then fixed for 10 min at room temperature with freshly prepared 3.6% paraformaldehyde, 0.024% saponin, 1 mM Na 3 VO 4 in PBS. Free aldehyde groups were quenched with 0.1% NaBH 4 for 10 min, and nonspecific sites were blocked with 1% BSA. To detect total EGFR, a mixture of anti-EGFR monoclonal antibodies was used (528, 13A9, 225; 10 g/ml each). To detect ⌬EGFR, D1133 monoclonal antibody was used at a concentration of 10 g/ml. All antibody and rinse solutions contained 1% BSA, 0.012% saponin, and 1 mM Na 3 VO 4 . Coverslips were incubated with antireceptor antibodies together with 10 g/ml affinity purified rabbit antiphosphotyrosine antibodies for 1 h followed by staining with fluorescein isothiocyanatelabeled goat anti-mouse and Texas Red-labeled goat anti-rabbit antibodies (1:100; Cappel Laboratories) for 45 min. The coverslips were mounted in ProLong antifade medium (Molecular Probes, Inc.) and viewed with a Nikon inverted fluorescence microscope with a 40 ϫ oil immersion objective. A multi-dye filter set was used in which excitation filters of 485 nm (fluorescein isothiocyanate) and 575 nm (Texas Red) were selected from a computer-controlled filter wheel in conjunction with a multi-wavelength emitter and dichroic filter set (460, 525, and 640 nm; XF56 set from Omega Optical). Fluorescein isothiocyanate and Texas Red images were acquired separately using a Photometrics cooled CCD camera and a Macintosh workstation running OncorImage software. Images were scaled to 256 gray levels using Adobe Photoshop 3.0 on the Macintosh and transferred to TMAX-100 film using an AFGA film recorder.
Rate of Endocytosis-The specific internalization rates of wild-type EGFR (wt EGFR) and ⌬EGFR were determined by measuring endocytosis of 125 I-labeled 13A9 monoclonal antibody. The antibody binds equally well to both the wt EGFR and ⌬EGFR and does not affect the binding of EGF to the wt EGFR. This allowed evaluation of receptor endocytosis in both the empty and occupied state. 13A9 antibody was iodinated with 125 I (Amersham Corp.) to a specific activity of 1500 cpm/fmol as described previously (35). Cells grown to confluency on fibronectin-coated plates were incubated at 0°C for 90 min with 4 nM 125 I-labeled 13A9 in the presence or absence of 1 g/ml EGF. The cells were rapidly warmed to 37°C by the addition of prewarmed medium. The relative amount of internalized versus surface-associated antibody was determined at 2-min intervals by acid stripping as described (9). The stripping efficiency was approximately 93%. The data were converted to internalization plots as described previously (9) and normalized to total surface binding of 13A9 (36).
In Vitro Kinase Activity-Cells expressing wild-type and mutant EGFR were washed in PBS containing 2 mM EDTA and 2 mM EGTA before being solubilized and homogenized at 4°C in buffer containing 10 mM Hepes, pH 7.9, 10 mM NaCl, 1% glycerol, 1% Triton X-100, 2 g/ml leupeptin, 2 g/ml aprotinin, 4 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride. Cell lysates were centrifuged at 12,000 ϫ g for 5 min at 4°C. Mutant receptor supernatants were incubated with a 1:50 dilution of monoclonal antibody D124, and wt EGFR supernatant was incubated with 1:100 dilution of monoclonal antibody, 13A9, at 4°C for 1 h. Protein A-Sepharose was then added for 30 min at 4°C. Immunoprecipitates were washed four times in buffer containing 10 mM Hepes, pH 7.9, 300 mM NaCl, 1% glycerol, 2 g/ml leupeptin, 4 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride. Immunoisolated wt EGFR was treated with 200 ng/ml EGF for 10 min at 22°C prior to start of the kinase assay. Kinase reaction mixtures contained 12.5 mM Hepes, pH 7.4, 10 mM MnCl 2 , 50 M Na 3 VO 4 , 10 M [␥-32 P]ATP (1.0 -2.8 ϫ 10 4 cpm/pmol), immunoprecipitated EGFR, and 100 M 1173 peptide (RRKGSTAENAEYLRV). Phosphorylation reactions proceeded for 10 min at 22°C and were terminated by the addition of trichloroacetic acid (final concentration, 6.1%). BSA (100 g) was added to facilitate the precipitation of EGFR, and the reactions were incubated at 4°C for 10 min. Precipitated receptors were centrifuged for 5 min at 12,000 ϫ g, and phosphate incorporation was determined by spotting the supernatant on phosphocellulose paper (Whatman P-81, 2.4-cm circle). The phosphocellulose was washed four times in 500 ml of 75 mM phosphoric acid for 15 min at 4°C, rinsed in acetone, and dried. The amount of retained radioactivity on the phosphocellulose was then measured. Protein concentrations were determined by separating receptors and protein standards on 7.5% SDS-polyacrylamide gels, trans-ferring to Immobilon, probing with the anti-EGFR polyclonal antibody, Ksm, followed by an alkaline phosphatase conjugated anti-rabbit antibody, and scanning the blots with a densitometer.
Tumorigenicity-For subcutaneous inoculation, U87MG cells (1 ϫ 10 6 ) and U87MG cells expressing mutant EGFR (1 ϫ 10 6 or 2 ϫ 10 5 ) were suspended in 0.1 ml of PBS and injected simultaneously into the left or right flanks, respectively, of 4-to 5-wk-old female nude mice of BALB/c background. The growing tumors were measured twice a week with a caliper, and tumor volumes were calculated using width (a) and length (b) measurements (a 2 b/2, where a Ͻ b) (37). For intracerebral stereotactic implantation, 1 ϫ 10 5 -5 ϫ 10 5 cells in 5 l of PBS were inoculated into the corpus striatum in the right hemisphere of the nude mouse brain (38). Brains were removed 12 days postimplantation, embedded in OCT compound (Miles), frozen in liquid nitrogen, and stored at Ϫ80°C. Thin cryostat sections (5-6 m) were stained with 528 antibody and counter-stained with hematoxylin. Tumors were microscopically measured, and the volumes of tumors were calculated as described above.

RESULTS
Defective Down-regulation of ⌬EGFR-The autophosphorylation of wt EGFR induced by ligand binding results in a conformational change proposed to expose cryptic endocytic and lysosomal targeting codes required for induced internalization and down-regulation (11,39). Kinase-inactive EGFR are defective in ligand-induced down-regulation (40,41), whereas deletion of the trafficking sequence codes in the carboxyl terminus of kinase-active EGFR results in ligand-dependent transformation and tumorigenesis by these attenuation-defective EGFR (12,42). The extent of autophosphorylation of mutant ⌬EGFR was much less than that of ligand-activated wt EGFR (26), raising the question of whether these mutant receptors can undergo the endocytosis and down-regulation characteristic of activated EGFR. Receptor distribution was first analyzed by immunofluorescence microscopy using monoclonal antibodies directed to the fusion junction in the ectodomain of ⌬EGFR, to wt EGFR, and to phosphotyrosine. As shown in Fig. 1, both wild-type and ⌬EGFR were predominantly located at the cell surface of U87MG cells (Fig. 1, a and e). Staining of permeabilized cells localized both wt EGFR and ⌬EGFR diffusely on the cell surface with enhancements at cell margins in agreement with previous reports (43,44). After addition of EGF, wt EGFR were lost from the cell surface and appeared in perinuclear vesicles corresponding to endosomes and lysosomes (Fig. 1c). In contrast, the distribution of ⌬EGFR was not changed upon addition of EGF, and receptors remained at the cell surface (Fig. 1g). In cells expressing wt EGFR, phosphotyrosine staining was most marked in focal adhesions (Fig. 1b) and, after addition of EGF, increased phosphotyrosine co-localized with internalized EGFR in perinuclear vesicles (Fig. 1d). In cells expressing ⌬EGFR, phosphotyrosine was detected primarily with a punctate distribution at the cell surface which corresponded to the distribution of ⌬EGFR (Fig. 1f); after addition of EGF this staining at the cell surface did not change (Fig. 1h).
To determine whether the phosphorylation detected in these assays could be primarily attributed to ⌬EGFR (Fig. 1, e-h), extracts from U87MG and U87MG.⌬EGFR cells were analyzed by Western blotting with anti-phosphotyrosine antibodies. Phosphorimage quantitation indicated that a single peak corresponding to ⌬EGFR accounted for the majority of the measurable increase in phosphotyrosine ( Fig. 2A). Identical results were obtained when these cells were chronically treated with EGF or with the antagonist monoclonal anti-EGFR antibody 225 (data not shown), confirming lack of effect of both ligand agonist and an antagonist monoclonal antibody on tyrosine kinase activity of ⌬EGFR. To confirm that the single peak corresponding to 140 -155 kDa was indeed ⌬EGFR, the mutant receptors were repeatedly immunodepleted from U87MG.⌬EGFR cell lysates (Fig. 2B, lanes 3-4) using a mono-clonal antibody specific to ⌬EGFR. In the ⌬EGFR-depleted lysate the overwhelming majority of phosphotyrosine immunoreactivity in the 140 -155-kDa region on Western blots had disappeared (Fig. 2B, lane 5), indicating that phosphotyrosine immunoreactivity detected at this region (Fig. 2B, lane 2) is ⌬EGFR.
Because the rate-limiting step in down-regulation is internalization from the cell surface (9,45), the rates of endocytosis of wt EGFR and ⌬EGFR were directly measured using the 13A9 monoclonal antibody that recognizes both. Previous studies have indicated that the rate of endocytosis of monoclonal antibodies bound to the ectodomain of wt EGFR accurately reflects basal and ligand-stimulated internalization kinetics (9). The rate of endocytosis of ⌬EGFR corresponded to the basal rate of endocytosis of non-activated and kinase-inactive EGFR (Fig. 3) (9) and was consistent with random entrapment in coated pits (9). In sharp contrast, ligand-activated wt EGFR exhibited a significantly higher rate of endocytosis. ⌬EGFR were thus defective in down-regulation at least in part due to a low rate of constitutive internalization equivalent to that of non-activated wt EGFR.
Essential Sites of Autophosphorylation in ⌬EGFR-The major phosphotyrosine containing protein in U87MG.⌬EGFR cells was the ⌬EGFR itself ( Figs. 1 and 2) suggesting that autophosphorylation was critical to biological signaling by this mutant receptor. To determine the contribution of the intrinsic tyrosine kinase activity and autophosphorylation of ⌬EGFR to tumorigenicity, several derivative ⌬EGFR mutants were constructed (Fig. 4A). The kinase activity of ⌬EGFR was inactivated by introducing a point mutation in the ATP-binding site which changed lysine 721 to methionine (8) (DK, Fig. 4A). Other ⌬EGFR cDNAs were constructed with point mutations that altered the tyrosine phosphorylation sites identified in wt EGFR (Tyr-992, Tyr-1068, Tyr-1086, Tyr-1148, and Tyr-1173) (46 -48), either singly (DY1, Y1173F) or in combinations (DY2-5). The constructs were introduced into U87MG cells individually by retroviral-mediated gene transfer and selected, and populations expressing levels of mutant receptors similar to that of U87MG.⌬EGFR were sorted and expanded for further analysis (Fig. 4A). As a control, U87MG cells infected with viruses carrying the wt EGFR gene and expressing a similar level of wt EGFR were also selected by the same procedure. The presence of equivalent amounts of mutant (140 -155 kDa) and wt receptors (170 kDa) in the cells was confirmed by probing  -d and g-h). The cells were fixed, permeabilized, and stained for EGFR or ⌬EGFR (left panels) and simultaneously for phosphotyrosine (right panels).

FIG. 2. Phosphotyrosine-containing proteins in cells expressing ⌬EGFR.
A, total protein from U87MG cells (a) or U87MG.⌬EGFR cells (b) was separated on 5-15% gradient gels, transferred to nitrocellulose, and probed with antiphosphotyrosine polyclonal antibodies and 125 I-Protein A. Dried blots were exposed to storage phosphor plates to approximately 50% dynamic range. Shown are the profiles from U87MG cells (a), U87MG.⌬EGFR cells (b), and the difference between the two (c). Western blots using anti-EGFR antibodies (Fig. 4B, lower  panel).
While ⌬EGFR was constitutively autophosphorylated (Fig.  4B, upper panel, lanes 3 and 4), DK EGFR was devoid of any significant tyrosine phosphorylation (Fig. 4B, upper panel, lane  5). Tyrosine phosphorylation of DK EGFR could not be restored by activation of the endogenous wt EGFR in the same cells; only a minor amount of DK EGFR phosphorylation was observed (Fig. 4B, upper panel, lane 6), indicating a lack of substantial cross-phosphorylation between ligand-activated wildtype and mutant receptors. This conclusion was supported by the lack of phosphorylation of wt EGFR which co-existed with the constitutively active ⌬EGFR in the same cells (Fig. 4B,  upper panel, lane 3). Autophosphorylation of mutant receptors was principally at tyrosine residues 1068, 1148, and 1173, because receptors mutated at these sites (DY3) showed almost complete loss of phosphorylation (Fig. 4B, upper panel, lanes 11  and 12). Mutants which retained one (DY2 and DY4) or two (DY1) of these residues were still phosphorylated (Fig. 4B,  upper panel, lanes 7-10, 13, and 14). Receptors with mutations of all five tyrosine residues (DY5) were somewhat phosphorylated, albeit at significantly reduced levels (Fig. 4B, upper  panel, lanes 15 and 16).
In order to quantify autophosphorylation of the various mutant EGFRs, receptors were immunoisolated before probing with anti-phosphotyrosine antibody. Fig. 5A shows that the extent of phosphorylation per molecule of ⌬EGFR was approximately 12% that of wt EGFR activated by EGF treatment. Analysis of the DY series of Tyr to Phe mutations indicated that Tyr-1173 appeared to be the major autophosphorylation site since a single mutation at this residue decreased phosphorylation to 35% that of ⌬EGFR. Furthermore, mutant receptor DY4, which retained only the Tyr-1173 site, was phosphorylated to 42% that of ⌬EGFR (Fig. 5B). Phosphorylation of mutant receptors with alterations in more than one autophosphorylation site was even lower, with decreases to 25 and 1% for DY2 and DY3, respectively (Fig. 5B). The phosphorylation level of DY5 was 3% that observed for ⌬EGFR, suggesting that mutation of all five tyrosines resulted in accessibility of the intrinsic kinase to some minor phosphotyrosine site. In contrast, mutation of the kinase domain completely abolished autophosphorylation of ⌬EGFR (Fig. 5B).
Autophosphorylation of EGFR tyrosine residues by ligand binding not only activates binding sites for signaling molecules but may also regulate the catalytic activity of the receptor by altering the conformation of the kinase domain and/or by changing the access of cellular substrates to the kinase active site (49,50). This raised the question of whether mutations of the tyrosine residues in ⌬EGFR affected the kinase activity. To evaluate the contribution of the kinase domain and the phosphorylation sites of ⌬EGFR in the enhanced tumorigenic effect, the kinase activities of ⌬EGFR containing mutations in tyrosine residues were determined. Cells overexpressing wild-type or mutant receptors were solubilized, and the receptors were immunoisolated using anti-wt EGFR or anti-⌬EGFR specific antibodies. Receptor kinase activity was assayed in vitro using the synthetic substrate, 1173 peptide (51). While point muta- B, Western blot analysis of wild-type and mutant EGFR and the degree of autophosphorylation. Cells expressing the indicated mutant or wildtype EGFR (U87MG.wtEGFR) were treated without (Ϫ) or with (ϩ) 100 ng/ml EGF for 5 min. 6 ϫ 10 4 cell equivalents of protein was loaded onto each lane, and proteins were separated on 7.5% SDS-polyacrylamide gels and transferred to nitrocellulose membranes. Tyrosine phosphorylation of the receptors was detected by anti-phosphotyrosine monoclonal antibody, PY-20 (upper panel). EGFR were subsequently identified on the same membrane by anti-EGFR monoclonal antibody (C13) (lower panel). Open arrow, wild-type EGFR; solid arrow, mutant EGFR. tion of Lys-721 inactivated the kinase activity of the DK receptor, mutations of tyrosine residues did not have a measurable effect on the in vitro kinase activity of the various DY receptors ( Table I).
Dependence of Tumor Growth on Autophosphorylation of ⌬EGFR-In order to confirm the tumorigenic importance of the intrinsic tyrosine kinase activity of ⌬EGFR, the ability of U87MG.DK cells to form tumors in nude mice was measured. U87MG.⌬EGFR cells grew markedly faster than U87MG cells when the same number of cells (1 ϫ 10 6 ) were implanted subcutaneously into the flanks of nude mice (Fig. 6A), in agreement with a previous report (26). Reducing the number of implanted U87MG.⌬EGFR cells 5-fold still resulted in a profound growth advantage of the mutant receptor containing cells when compared with parental U87MG cells. U87MG.DK cells which expressed the same level of DK receptors as that of ⌬EGFR in U87MG.⌬EGFR cells showed a similarly low tumorigenic potential as parental U87MG cells (Fig. 6A), indicating that inactivation of the intrinsic tyrosine kinase activity abolished the enhancement of tumorigenicity mediated by ⌬EGFR.
To determine whether the loss of growth advantage of DK receptor containing U87MG cells was specific to the subcutaneous site, these glioblastoma cells were stereotactically implanted into the brains of nude mice. Animals which received 1 ϫ 10 5 U87MG.⌬EGFR cells had tumor masses of approximately 60 mm 3 12 days postimplantation which were more than 100-fold larger than tumors from the inoculation of 5 ϫ 10 5 U87MG.DK or parental U87MG cells (Fig. 7, a-c). Western blotting analysis showed that tumors obtained from U87MG.DK implantations expressed endogenous EGFR of 170 kDa as well as smaller species of 140 -155 kDa corresponding to the DK receptor. These DK receptors were expressed at a level comparable with that of U87MG.DK cells grown in vitro FIG. 5. Quantitation of tyrosine phosphorylation of mutant EGFR. A, U87MG.wtEGFR cells and U87MG.⌬EGFR cells were treated without (Ϫ) or with (ϩ) 100 ng/ml EGF for 5 min at room temperature prior to cell lysis. Mutant and wild-type EGFR were immunoisolated by anti-EGFR antibody, 528. Receptors loaded in each lane were isolated from 1 ϫ 10 5 cells (wtEGFR lanes) or 3 ϫ 10 5 cells (⌬EGFR lanes). Receptor tyrosine phosphorylation was detected by Western blot analysis as described in Fig. 4B, and autoradiograms were quantitated by densitometry. Tyrosine phosphorylation per wild-type receptor (wtEGFR lanes) or per mutant receptor (⌬EGFR lanes) is shown below each and calculated as a percentage of wild-type EGFR (100%). B, characterization of ⌬EGFR autophosphorylation by mutational analysis. Receptors loaded in each lane were immunoisolated by the mutant receptor specific monoclonal antibody D1133 from 6 ϫ 10 4 cells. Receptor tyrosine phosphorylation was detected by Western blot analysis as described in Fig. 4B and quantified by densitometry. Tyrosine phosphorylation per mutant receptor is shown below each lane and calculated as a percentage of ⌬EGFR (100%).

TABLE I
In vitro kinase activity of wild-type and mutant EGFR U87MG cells expressing wild-type and mutant EGFR were solubilized and receptors were immunoisolated using the anti-⌬EGFR specific monoclonal antibody D124 and the anti-EGFR monoclonal antibody 13A9. Kinase activity was measured by adding peptide substrate, [␥-32 p]ATP, and MnCl 2 at 22°C as described under "Experimental Procedures." wt EGFR was treated with EGF for 10 min at 22°C, prior to the kinase assay. Protein concentrations were determined by Western blotting using Ksm, an anti-EGFR antibody, and blots were quantitated using densitometry. Purified EGFR was used as a standard. n ϭ 3 for each mutant EGFR data point.   , n ϭ 4), or U87MG.wtEGFR cells (૾, n ϭ 4) were inoculated into the right flank of nude mice while 1 ϫ 10 6 U87MG cells (j, n ϭ 4 (A) or n ϭ 28 (B)) were injected into the left flank of the same animals. Tumor size was quantitated at the indicated times. Similar results were obtained in two to three additional experiments. and were not tyrosine-phosphorylated (data not shown), demonstrating lack of activity and stable expression of kinaseinactive DK EGFR in vivo. Taken together, these data indicate that the intrinsic kinase activity is essential for the enhanced tumorigenicity mediated by ⌬EGFR.
Mutations of the autophosphorylation sites significantly affected the ability of ⌬EGFR to enhance tumor formation in nude mice. The kinetics of subcutaneous tumor growth shown in Fig. 6B indicate that mutation of any of the autophosphorylation sites, singly or in combination, dramatically reduced the enhanced tumorigenic activity conferred by the ⌬EGFR in U87MG cells. Cells bearing these phosphorylation site mutants, with the exception of DY3, grew only slightly faster than the same number (1 ϫ 10 6 ) of implanted parental U87MG cells, and much more slowly than the tumors resulting from 5-fold fewer inoculated U87MG.⌬EGFR cells. U87MG.DY3 cells repeatedly formed tumors more slowly than the parental cells, and the expression of the DY3 mutant receptor was not stable either in tissue culture or in tumors. Similar results were obtained when these cells were injected into the brains of nude mice (Fig. 7, d-h). The average volume of tumors derived from cells expressing ⌬EGFR with tyrosine autophosphorylation site mutations was 30 -600-fold less than that of tumors formed from 5-fold fewer U87MG.⌬EGFR cells. These data strongly suggest that the tumorigenic effect of ⌬EGFR was mediated through phosphorylation of its carboxyl-terminal tyrosine residues. Overexpression of wt EGFR did not confer significantly enhanced tumorigenic activity in U87MG cells in either subcutaneous or brain implantations ( Fig. 6B and Fig. 7i), indicating that ligand-dependent wt EGFR functioned less effectively than constitutively active ⌬EGFR in the animal model.

DISCUSSION
Deletion of amino acid residues 6 -274 located in the ectodomain of EGFR creates a highly tumorigenic receptor tyrosine kinase as measured by growth of subcutaneously and intracerebrally implanted glioma expressor cells (26). Because this mutant receptor occurs in a variety of human tumors (13)(14)(15), the basis for its enhanced proliferative effects are of considerable interest. The ⌬EGFR has several features in common with v-erbB, the principal oncogene of avian erythroblastosis virus (52), including failure to bind ligand and constitutive activation of intrinsic protein tyrosine kinase activity as assessed by autophosphorylation (53,54). v-erbB resulted from retroviral insertion that reproducibly deleted most of the ectodomain of the avian EGFR (55), whereas ⌬EGFR results from a common gene deletion that occurs naturally during malignant progression (18,21,24,56). However, the transforming activity and tissue specificity of v-erbB is dependent on additional truncation and/or mutation of the carboxyl terminus (57,58), raising the question of whether a small deletion of the ligand-binding domain in ⌬EGFR alone is capable of eliciting any biological effect. The present studies confirm the hypothesis that the constitutive tyrosine kinase activity resulting from ectodomain deletion is essential to tumorigenic effects in human tumor cells, where this mutant receptor naturally occurs. Since the extent of autophosphorylation of ⌬EGFR was small compared with ligand-activated wild-type EGFR, the basis for strong mitogenic signaling was uncertain. The present studies indicate that enhanced tumorigenesis of ⌬EGFR is strictly dependent on intact autophosphorylation sites located in the regulatory carboxyl terminus. Our observation that ⌬EGFR are defective in endocytosis and down-regulation suggests that the effects of the small extent of constitutive activation are greatly amplified due to failure of normal attenuation mechanisms.
Individual autophosphorylation sites in several tyrosine kinase receptors, an event particularly well studied in PDGF receptors, exhibit high specificity for binding SH2 and PTB domains (59); mutation of individual sites abrogates signaling via particular proteins (60). Autophosphorylation sites of EGFR appear to exhibit less specificity (2), although heterodimerization with and transphosphorylation of molecules such as erbB2 and erbB3 may provide coupling to specific molecules such as phosphatidylinositol 3-kinase (61). In ⌬EGFR, Tyr-1173, the favored autophosphorylation site in ligand-activated wt EGFR (46), appears to be the major site of autophosphorylation. Mutation of Tyr-1173 reduced its autophosphorylation by ϳ65%, and mutation of all known autophosphorylation sites except Tyr-1173 yielded a receptor with ϳ45% the extent of autophosphorylation. Mutation of the three major autophosphorylation sites at Tyr-1068, Tyr-1148, and  , data not shown) or in combination (DY3, DY4, DY5) abolished the enhanced tumorigenesis characteristic of ⌬EGFR. Since these measurements reflect the average extent of phosphorylation per receptor, these values may reflect heterogeneity in the phosphorylation of any one receptor. The requirement that all major sites of tyrosine autophosphorylation be intact for enhanced tumorigenesis could reflect a combinatorial role for interaction of multiple SH2 and phosphotyrosine binding domain proteins with constitutively activated ⌬EGFR. Alternatively, given the finding that ⌬EGFR are autophosphorylated to only ϳ10% the level of ligand-activated wild-type receptors, all autophosphorylation sites may be required to reach a critical threshold. These characteristics distinguish ⌬EGFR from wt EGFR where it is necessary to mutate multiple tyrosine autophosphorylation sites to block ligand-dependent transformation (3).
There is evidence that ⌬EGFR couple to the mitogen-activated protein kinase signaling pathway utilized by wt EGFR (62) but more weakly (63). Montgomery et al. (63) observed activation of mitogen-activated protein kinase/extracellular signal-regulated kinase kinase and mitogen-activated protein kinases in NIH 3T3 cells expressing ⌬EGFR. The extent of activation was modest presumably due to enhanced tyrosine phosphatase activity. Further studies confirmed minimal activation of mitogen-activated protein kinases but could not detect increases in GTP-bound Ras (64). However, Prigent et al. While ⌬EGFR may couple to additional signaling pathways, it appears to use the same mitogenic signal transduction pathways as wt EGFR consistent with the low level constitutive activation of the mutant receptor. Moreover, it has been reported (64) that ⌬EGFR is constitutively dimerized. We noted a small (Ͻ5%) extent of dimerization of ⌬EGFR that was unaffected by addition of ligand. 3 The extent of dimerization observed was roughly proportional to the extent of activation of ⌬EGFR (ϳ10%) and efficacy of cross-linking. ⌬EGFR thus exhibits neither strong homo-nor hetero-dimerization.
Autophosphorylation of EGFR provides sites both for assembly of components of the signal transduction cascade and for exposure of cryptic sequences in the carboxyl terminus necessary for ligand-induced endocytosis and down-regulation. In the absence of the carboxyl terminus, ligand-activated EGFR exhibit enhanced tumorigenesis due to failure to increase both ligand and receptor internalization (12,42,(67)(68)(69). Ectodomain-deleted ⌬EGFR are also defective in endocytosis and down-regulation. A previous report that ⌬EGFR localized within cells rather than on the cell surface (70) was not confirmed in the present studies nor in similar studies using several other glioblastoma cell lines (71). 4 Instead ⌬EGFR expressed in brain tumor cells are mainly localized to the cell surface reflecting a low constitutive rate of endocytosis relative to the normal rate of receptor recycling (35). These constitutively active receptors thus escape the major mechanism used to attenuate receptor signaling. Although Moscatello et al. (64) recently reported that 53% of transfected ⌬EGFR expressed in NIH3T3 cells were located on the cell surface, they also showed 59% of the wt EGFR expressed in the same cell type was on the cell surface, indicating that the distribution of the EGFR may be cell type-specific. Indeed, we have examined the relative distribution of wt EGFR and ⌬EGFR in both the NIH3T3 variant NR6, which lacks endogenous EGFR, and the human mammary epithelial cell line 184A1, which expresses high levels of endogenous EGFR. The distribution of wt EGFR was different between NR6 and 184A1 cells, being almost exclusively cell-surface localized in NR6 cells and distributed between surface and intracellular compartments in 184A1 cells. The addition of EGF shifted the wt EGFR to an intracellular compartment in both cell types. 5 Significantly, the distribution of the ⌬EGFR was the same as unstimulated wt EGFR in both cell types and was not altered by addition of EGF. These results suggest that although the ⌬EGFR displays tyrosine kinase activity, it may interact with the cellular trafficking machinery in the same way as empty wt EGFR.
The constitutive kinase activation, autophosphorylation levels set near the threshold for enhanced mitogenesis, and the defect in endocytosis may all reflect partial activation of ⌬EGFR. Conformational changes sufficient to allow autophosphorylation to ϳ10% is sufficient for receptor mitogenic signaling but insufficient for exposure of endocytic sequence codes necessary for down-regulation and attenuation of signal transduction. The recent observation that the kinase activities of wt EGFR and ⌬EGFR can be differentially blocked by various tyrosine kinase inhibitors is consistent with the idea that the conformational changes induced by ectodomain deletion in ⌬EGFR differs from ligand-activated wt EGFR (72). We hypothesize that deletion of exons 2-7 removes inhibitory constraints sufficient to allow low level constitutive activation of the intrinsic protein tyrosine kinase activity of EGFR. Although specific autophosphorylation sites may engage specific targets, removal of any one site abolishes the tumorigenic effect of the mutant receptor. Since the level of constitutive activation appears near the threshold for mitogenic responses, we propose that failure to internalize ⌬EGFR is a critical defect allowing inappropriate growth signals. Persistence of ⌬EGFR at the cell surface, the apparent site of assembly of many signaling complexes, would prolong and enhance its low level of activity. ⌬EGFR is thus seen as active for mitogenic signaling but escapes detection by the endocytic trafficking system so that signaling is unattenuated.