Domain-specific Interactions between the p185 neu and Epidermal Growth Factor Receptor Kinases Determine Differential Signaling Outcomes*

We expressed the epidermal growth factor receptor (EGFR) along with mutant p185 neu proteins containing the rat transmembrane point mutation. The work concerned the study of the contributions made by various p185 neu subdomains to signaling induced by a heterodimeric ErbB complex. Co-expression of full-length EGFR and oncogenic p185 neu receptors resulted in an increased EGF-induced phosphotyrosine content of p185 neu , increased cell proliferation to limiting concentrations of EGF, and increases in both EGF-induced MAPK and phosphatidylinositol 3-kinase (PI 3-kinase) activation. Intracellular domain-deleted p185 neu receptors (T691stop neu) were able to associate with full-length EGFR, but induced antagonistic effects on EGF-dependent EGF receptor down-regulation, cell proliferation, and activation of MAPK and PI 3-kinase pathways. Ectodomain-deleted p185 neu proteins (TΔ5) were unable to physically associate with EGFR, and extracellular domain-deleted p185 neu forms failed to augment activation of MAPK and PI 3-kinase in response to EGF. Association of EGFR with a carboxyl-terminally truncated p185 neu mutant (TAPstop) form did not increase transforming efficiency and phosphotyrosine content of the TAPstop species, and proliferation of EGFR·TAPstop-co-expressing cells in response to EGF was similar to cells containing EGFR only. Thus, neither cooperative nor inhibitory effects were observed in cell lines co-expressing either TΔ5 or TAPstop mutant proteins. Unlike the formation of potent homodimer assemblies composed of oncogenic p185 neu , the induction of signaling from p185 neu ·EGFR heteroreceptor assemblies requires the ectodomain for ligand-dependent physical association and intracellular domain contacts for efficient intermolecular kinase activation.

The transmembrane mutation in rat p185 neu (also termed Tneu) (12) serves as a paradigm for receptor dimerization that leads to constitutive kinase activation contributing to oncogenic transformation (11)(12)(13). Additional support for this mechanism has come from the identification of a naturally occurring activated EGFR oncoprotein (⌬EGFR or EGFRvIII) in human tumors, which forms constitutive dimers and confers increased tumorigenicity (14,15). Gene amplification and overexpression of ErbB-2 have been observed in a high frequency of human adenocarcinomas, including those of the breast and ovary, and these features correlate with poor clinical prognosis (16,17). Experimental support for this model is provided by in vitro transformation assays using cell lines overexpressing either protooncogenic rat p185 c-neu or human ErbB-2 at levels of 10 6 receptors/cell (18,19). Biochemical and biophysical analysis of baculovirus-expressed p185 neu proteins further support the notion of receptor oligomerization as a mechanism of kinase activation of normal holoreceptors (20,21).
Heterodimeric interactions govern many signaling properties within the ErbB receptor family. Co-expression of EGFR and p185 c-neu at modestly elevated levels (10 5 /cell) (but not either receptor independently) results in synergistic transformation (22), due to increase of the ligand binding affinity and catalytic kinase activity (23,24). Heterodimerization of EGFR and ErbB-2 has also been observed in human breast tumor lines (25). Moreover, ligand treatment promotes the assembly of an activated p185 c-neu ⅐EGFR kinase complex in many cells (24), resulting in novel distinct cellular signaling events (26). Therefore, the receptor tyrosine kinase ensemble can be activated not only by homodimer formation, but also by heterodimeric associations. In this regard, endodomain interactions between p185 neu and EGFR appear to influence functional signaling outcomes (27).
In response to EGF or Neu differentiating factor/heregulin (a ligand for ErbB-3 and ErbB-4) family ligands (28,29), EGFR and ErbB-2 both form heterodimers with ErbB-3 and ErbB-4 (30 -34). Heterodimers between p185 neu ⅐ErbB-2 and ErbB-3 are associated with activated signaling and the transformed phenotype in primary human cancer cells (35). Existence of an ErbB-3⅐ErbB-4 heterodimer has not been convincingly demonstrated to date. More recent data support the notion that p185 neu ⅐ErbB-2 is the preferred heterodimerization partner of all ErbB receptors and a mediator for divergent cellular signaling in many distinct cell types (34,36).
The structural basis for ErbB receptor heterodimerization has not been completely defined and crystallographic information on dimerized ErbB receptor kinases is currently unavailable. Previous work has revealed that ectodomain interactions are sufficient to stabilize dimer formation between p185 neu and EGFR in fibroblasts and transformed cells (5,37,38), which is supported by observations showing that a partial deletion of the EGF receptor ectodomain still allow dimer formation and receptor activation (14,15). Although the transmembrane alone can stabilize the formation of p185 neu homodimers, the relative contributions of the transmembrane region and the ectodomain have not been directly compared regarding the formation of signaling heterodimers.
In this study, we have constructed various p185 neu deletion mutants in order to specifically compare signaling events resulting from associations between EGF receptors and either p185 neu ectodomain-or endodomain-derived mutant receptors. We have co-expressed EGFR with low levels of p185 neu proteins, or their mutant derivatives, to monitor p185 neu -mediated enhancement of cell growth and transformation in vitro and in vivo, and to analyze the influence of EGF-induced heterodimeric receptor interactions on downstream signaling effectors. Signaling resulting from heterodimeric associations between full-length EGFR and mutant p185 neu proteins has revealed the functional importance of p185 neu subdomains in the induction of Ras/extracellular signal-regulated kinase (ERK) and phosphatidylinositol 3-kinase (PI 3-kinase) pathways contributing to cell growth and transformation.

EXPERIMENTAL PROCEDURES
Antibodies-As described previously (20,39,40), monoclonal antibody 7.16.4, polyclonal antiserum ␣-Bacneu, and NCT are reactive with the ectodomain, intracellular domain, and carboxyl terminus of p185 neu , respectively. mAb 225 reactive with the ectodomain of EGFR was obtained from Dr. John Mendelsohn (M. D. Anderson Cancer Center, Dallas, TX). A polyclonal rabbit antiserum specifically against the COOH terminus of EGFR (termed CT) was provided by Dr. Stuart Decker (40). The anti-phosphotyrosine monoclonal antibody, PY20, was obtained from Santa Cruz Biotechnology (Santa Cruz, CA.).
DNA Constructs-All the deletion mutants were derived from the rat oncogenic p185 neu cDNA containing a single point mutation (V664G) in the transmembrane region. The TAPstop mutant, containing a 122-aa truncation of the COOH terminus was prepared as described previously (41). A T691stop species was prepared by site-directed mutagenesis and substitution of a stop codon for Thr-691, resulting in a large cytoplasmic deletion (42,43). The ectodomain-deleted mutant T⌬5 neu protein was described previously (27). These cDNAs encoding for mutant p185 neu forms were all cloned into the pSV2neo r /DHFR vector as described (44) for expression in murine fibroblasts. These wild-type or mutant p185 neu cDNAs were also subcloned into pcDNA3 vector for transient expression in COS7 cells. pSR␣EGFR/hyg r vector (44) was used for full-length EGFR expression.
Transfection and Maintenance of Cell Lines-Ten micrograms of the p185 neu constructs were transfected into NR6 cells, a mouse fibroblast cell line devoid of endogenous EGF receptors (43), or NE91 cells expressing human EGFR (37) by calcium phosphate precipitation. After 2-3 weeks of selection with Geneticin (0.9 mg/ml), the established stable clones were screened and characterized. Gene amplification by methotrexate was used to increase the p185 neu receptor level. Expres-sion of p185 neu and its derivatives in resultant subclones was examined by flow cytometric analysis following anti-p185 neu mAb 7.16.4 staining. Surface expression of p185 neu proteins was then estimated by comparing the mean channel fluorescent intensity with that of B104-1-1 cells, as the level of p185 neu in B104-1-1 cells was previously determined by 125 I-labeled anti-neu mAb binding assay (22). EGFR numbers in NE91 cells and mutant p185 neu co-transfected cells were determined by Scatchard assays as described (37). These transfected clones were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 5% fetal bovine serum (FBS, HyClone) at 37°C in a 5% CO 2 atmosphere.
Cross-linking, Immunoprecipitation, and Immunoblotting Procedures-Subconfluent cells in 10-cm dishes were washed and starved in cysteine-free DMEM for 1 h, and grown in low cysteine-containing 5% FBS-DMEM containing 55 Ci/ml [ 35 S]cysteine (Amersham Pharmacia Biotech) for 16 h for metabolic labeling. Alternatively, the unlabeled cells were cultured overnight in 10-cm Petri dishes. After treatment with or without EGF, cells were washed twice with cold phosphatebuffered saline (PBS) and treated with PBS containing 2 mM membrane-impermeable cross-linker bis(sulfosuccinimidyl) suberate (BS 3 , Pierce), for 30 min. After quenching the cross-linking reaction with a buffer containing 10 mM Tris-HCl (pH 7.6), 0.9% NaCl, and 0.1 M glycine, cells were washed twice with cold PBS and solubilized with PI/RIPA buffer as described (24). The immunocomplexes were washed and solubilized, then separated by gradient SDS-PAGE gels (4 -7.5%). Proteins from metabolically labeled cells were analyzed by autoradiography. Proteins from unlabeled cells were transferred onto nitrocellulose and then immunoblotted with anti-phosphotyrosine mAb PY20, anti-EGFR CT, or anti-p185 antiserum as indicated in the figures. The protein signals were identified by the binding of 125 I-labeled protein A (NEN Life Science Products), or by enhanced chemiluminase (ECL) using ECL kit from Amersham Pharmacia Biotech.
Receptor Down-regulation Studies-Cells (1 ϫ 10 5 ) were plated in a six-well dish with DMEM containing 5% FBS overnight. Cells were then treated with EGF (50 ng/ml) for 0 -4 h and were harvested and washed with cold PBS containing 0.5% bovine serum albumin and 0.1% sodium azide. Cell preparations were then incubated with a saturating amount (0.5 g/reaction) of anti-neu mAb 7.16.4 or anti-EGFR mAb 225, or an irrelevant mAB (such as 9BG5 against the hemagglutinin of reovirus receptor), at 4°C for 30 min, restained with fluorescein isothiocyanate-conjugated anti-mouse IgG (Sigma) for another 30 min after extensive washing. Cells were then fixed with 2% paraformaldehyde and analyzed by flow cytometry (FACScan, Becton Dickinson), as described previously (37). Briefly, after subtracting the nonspecific background staining with 9BG5, the mean channel values from each time point were used to determine the percentage of surface expression of EGFR or p185 neu proteins at the various time points after EGF treatment.
In Vitro and in Vivo Transformation Assays-Anchorage-independent growth ability was determined by assessing the colony forming efficiency of cells suspended in soft agar (15,37). Cells (1000/dish) were suspended in 7% FBS-DMEM containing 0.18% agarose, and plated on 0.25% basal agar in each dish. Cells were fed with DMEM supplemented with 7% FBS-DMEM, 20 mM HEPES (pH 7.5). Colonies (Ͼ0.3 mm) were visualized at day 21 for all cell lines after stained with p-iodonitrotetrazolium violet (1 mg/ml). Each cell line was examined in triplicate samples for separate experiments.
To analyze the tumor growth in athymic mice, cells (1 ϫ 10 6 ) of each line were suspended in 0.1 ml of PBS and injected intradermally in the mid-dorsum of NCR nude mice. PBS alone was also injected as a control. Animals used in this study were maintained in accordance with the guidelines of the Committee on Animals of the University of Pennsylvania and those prepared by the Committee on Care and Use of Laboratory Animals of the Institute of Laboratory Animal Resource. Tumor growth was monitored twice a week up to 10 weeks. Tumor size was calculated by this formula: 3.14/6 ϫ (length ϫ width ϫ thickness) as described (27).
EGF-dependent Cell Proliferation Assay-The 3,(4,5-dimethylthiazol-2-yl)2,5-diphenyl tetrazolium bromide) (MTT) assay for measuring cell growth has been described previously (38). Briefly, cells (3000/well) of each cell line were seeded in 96-well plates overnight in DMEM containing 5% FBS. Cells were starved in serum-free ITS-DMEM for 48 h, then cultured in 100 l of the same medium plus various concentrations of EGF for another 48 h. 25 l of MTT solution (5 g/ml in PBS) were added to each well, and after 2 h of incubation at 37°C, 100 l of the extraction buffer (20% w/v SDS, 50% N,N-dimethyl formamide, pH 4.7) was added. After an overnight incubation at 37°C, the optical density at 600 nm was measured using an enzyme-linked immunosorbent assay reader. Each value represents a mean of four samples.
MAP Kinase and PI 3-Kinase Immune Complex Kinase Assays-COS7 cells were transiently transfected with pcDNA3-HA-ERK2 (a gift from Silvio Gutkind, National Institutes of Health, Bethesda, MD) and pSR␣EGFR/hyg r , along with either empty vector or plasmids expressing wild-type or mutant p185 c-neu using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions and assayed 48 h after transfection. Cells deprived of serum for 16 -20 h were treated with or without EGF (50 ng/ml) for 5 min. For MAP kinase assay, cells were lysed with RIPA buffer (25 mM Tris-HCl (pH 7.5), 0.3 M NaCl, 1.5 mM MgCl 2 , 1 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 20 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, 10 g/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin). Protein concentrations were determined by the BCA kit (Pierce). Equal amounts of protein (100 g) from cell extracts were immunoprecipitated with anti-HA (BabCo). After washing extensively, the immunocomplexes were then incubated with 50 l of reaction buffer (30 mM HEPES (pH 7.4), 10 mM MaCl 2 , 1 mM dithiothreitol, 5 M ATP) containing 1 Ci of [␥-32 P]ATP (NEN Life Science Products) and 2 g of myelin basic protein (Upstate Biotechnology Inc.). After incubation for 20 min at 30°C, kinase reactions were terminated by the addition of 2ϫ Laemmli sample buffer. The samples were then resolved by SDS-PAGE, and the phosphorylated myelin basic protein was visualized by autoradiography.

Expression of EGFR and/or Mutant p185 neu Proteins-Cell
lines expressing EGFR and various p185 neu deletion mutant proteins derived from full-length transforming p185 neu were all generated in the NR6 cell background (43). In addition, stable transfectants derived from NR6 fibroblasts expressing human EGFR (termed NE91 cells) were also generated. NE91 cells, as well as NR6 parental cells, were then transfected with various p185 neu cDNA constructs to express one of the following mutant p185 neu proteins with or without EGFR, respectively ( Fig.  1): (a) Er/p185 neu or p185 neu (full-length oncogenic p185 neu product), (b) Er/T691stop or T691stop (lacking 591 aa from the carboxyl terminus), (c) Er/TAPstop or TAPstop (a 122-aa truncation at carboxyl terminus), and (d) Er/T⌬5 or T⌬5 (an ectodomain deleted p185 neu product, also termed T⌬5). A schematic representation of the oncogenic p185 neu protein and its mutant derivative species is shown in Fig. 1.
B104-1-1 murine fibroblasts transformed by the expression of oncogenic p185 neu were used as a positive control, since surface expression of p185 neu, biochemical features of p185 neu homodimerization and p185 neu transforming potency have been characterized previously (13,22,27). As shown in Table I, relative expression levels of various p185 neu mutant proteins in selected clones were estimated by a comparison with B104-1-1 cells, while the expression of EGFR in these cells was estimated by Scatchard analysis. In order to observe an enhancement of EGFR-mediated cellular signaling and transformation, clone Er/p185 neu expressing a moderately low level of both receptors (ϳ10 4 /cell) was chosen. In other subclones, the expression of EGFR and/or mutant p185 neu proteins was approximately ϳ10 5 receptors/cell.
The Ectodomain of p185 neu Is Required for Heterodimerization with EGFR-Stable cell lines expressing EGFR and/or mutant p185 neu proteins were used to assess dimer formation using the chemical cross-linker BS 3 . As shown in Fig. 2, B104-1-1 cells expressing oncogenic p185 neu contained p185 neu homodimers (ϳ370 kDa) independent of ligand stimulation ( Fig.  2A, lane 1), due to the activating transmembrane mutation (12). A cell line expressing the ectodomain-derived T691stop neu alone was used as a control to demonstrate the sizes of the monomer and dimer of this truncated p185 neu protein, which migrated at approximately 115 kDa ( Fig. 2A, lanes 2 and 3), and at ϳ230 kDa in the presence of a chemical cross-linker ( Fig. 2A, lane 3).
In the presence of EGF, the 170-kDa monomeric form and the 340-kDa homodimer of EGFR were both detected in NE91 cells expressing EGFR alone, and in Er/T691stop cells ( Fig. 2A,  lanes 4 and 5, respectively). An additional intermediate band of ϳ285 kDa representing the heterodimer of EGFR and T691stop was clearly detectable upon anti-EGFR immunoprecipitation ( Fig. 2A, lane 5). The 285-kDa intermediate complex was similar to the heterodimer composed of EGFR and truncated N691stop derived from proto-oncogenic p185 c-neu as described previously (44), except that the heterodimeric EGFR⅐N691stop complex was even more predominant than the EGFR homodimer in those studies. Notably, T691stop is still able to complex with EGFR (lane 5) even under conditions favorable for T691stop homodimerization (lane 3). Densitometric analysis suggested that at least 50% of the EGFR associated with T691stop neu in a heterodimeric complex in Er/T691stop cells ( Fig. 2A, lane 5), further suggesting the strong preference for EGFR⅐p185 neu heterodimerization.
We have previously studied complex formation between the p185 neu and EGFR holoreceptors (22, 24, 37) and heterodimer- Tneu is the full-length transforming rat p185 neu . T691stop contains a stop codon substituting for Thr-691 at the amino terminus to the TK domain. TAPstop contains a 120-aa truncation within the carboxyl terminus of p185 neu . T⌬5 is generated by the deletion of ectodomain of p185 neu but retains ϳ10 aa and the signal sequence. These mutant p185 neu proteins were either expressed alone or co-expressed with EGFR in NR6 transfected cells.
ization between ectodomain p185 neu and either full-length (37, 44) EGFR or a form of EGFR that lacks the majority of subdomains 1 and 2 (15) by immunoprecipitation and immunoblotting using anti-receptor specific antibodies following EGF and chemical cross-linker treatment.
In this next set of studies, we extended these observations to novel species of p185 neu . Er/Nneu cells expressing higher levels of EGFR and normal p185 c-neu served as a positive control to examine the physical association of EGFR with truncated mutant p185 neu receptor forms (Fig. 2C, lane 1). The heterodimer between full-length p185 neu with EGFR in Er/neu cells could not be detected due to low expression levels of each receptor (data not shown). Abundant levels of the EGFR monomer and dimer were detected in Er/TAPstop cells by anti-EGFR immunoprecipitation and immunoblotting (Fig. 2C, lane 3). Analysis of anti-p185 neu immunoprecipitates by immunoblotting with anti-EGFR antisera indicated EGF-induced heterodimerization of EGFR and TAPstop in Er/TAPstop cells (Fig. 2C, lane 2). As expected, the size of this complex was slightly smaller when compared with the heterodimer of EGFR and full-length normal p185 neu in Er/Nneu cells (Fig. 2C, lane 1). The control cell line expressing TAPstop alone showed that TAPstop was only recognized by an anti-neu antibody (Fig. 2C, lane 5), but not by anti-EGFR antibody CT (Fig. 2C, lane 4).
Heterodimerization between EGFR and the ectodomain-deleted T⌬5 p185 neu mutant was also analyzed. T⌬5 can be recognized by either the ␣-Bacneu or anti-NCT polyclonal antisera reactive with the intracellular domain or carboxyl terminus of the p185 neu protein, respectively. Immunoblotting showed that the size of the T⌬5 neu mutant was approximately 95-97 kDa, and the detectable dimeric form was about ϳ200 kDa (Fig. 2B, lane 1). Er/T⌬5 cells express a high level of EGFR and T⌬5, as homodimers of either form were clearly detected in the presence of cross-linker (Fig. 2B, lanes 2 and 3), when compared with control cell lines NE91 and T⌬5 (Fig. 2B, lanes 1 and 4). However, unlike Er/T691stop and Er/TAPstop cells, the heterodimer between EGFR and T⌬5 in Er/T⌬5 cells was undetectable following EGF and BS 3 treatment since the predicted intermediate size (ϳ270 kDa) complex representing EGFR and T⌬5 heterodimer was not observed (Fig. 2B, lanes 2 and 3). In an attempt to identify the association of EGFR with this ectodomain-deleted T⌬5 protein, several alternative assays were performed, such as using the membrane-permeable chemical cross-linker DSP (Pierce), or a mild detergent digitonin lysis buffer. These methods were sensitive enough to detect the complex formation between full-length p185 neu and T⌬5 (27).

TABLE I Transformation parameters and relative receptor expression levels of cell lines
The number of EGFR on NE91 and other transfected cells was determined by Scatchard assays. Cell surface expression of neu proteins were estimated by comparing the mean channel fluorescent intensity with that from B104 -1-1 cells using flow cytometry analysis. p185 neu on B104 -1-1 cells was originally determined by an 125 I-labeled anti-neu mAb binding assay (22). For the tumor growth assay, individual clones (1 ϫ 10 6 cells/site) were injected intradermally into athymic mice. NT, no tumor after 10 weeks; ND, not determined.  However, the association of EGFR and T⌬5 was still undetectable (data not shown). Taken together, these results strongly suggest that the ectodomain of the p185 neu receptor is necessary and sufficient for heterodimerization with holoreceptor EGFR.
Tyrosine Kinase Activity in Living Cells-It has been well documented that EGF, in an EGFR-dependent manner, stimulated phosphorylation of the p185 c-neu and c-ErbB-2 gene products with a concomitant increase in their tyrosine kinase activities (46 -49). Heterodimerization of p185 and EGFR facilitates cross-phosphorylation (24,25), since a full-length, kinase-deficient p185 neu mutant (K757M) is trans-phosphorylated upon physical association with EGFR (37). We next examined the tyrosine phosphorylation level of p185 neu derivatives in living cells in response to EGF treatment. After the addition of EGF, oncogenic p185 neu and its derivatives were immunoprecipitated by anti-neu antibodies, and receptor phosphotyrosine content in vivo was detected by immunoblotting with an anti-phosphotyrosine antibody (PY20) (Fig. 3). Fulllength p185 neu from control B104-1-1 fibroblasts displayed constitutive kinase activity (Fig. 3A, lane 1). Upon EGF stimulation, there was indeed an additional increase in tyrosine kinase activity of p185 neu in Er/neu cells expressing lower amounts of the p185 neu protein (Fig. 3A, lanes 4 and 5), but not in cells expressing p185 neu alone (lanes 2 and 3). A weak tyrosine phosphorylation signal was detected in TAPstop cells (Fig. 3A,  lane 6). EGF stimulation did not appreciably increase the tyrosine phosphorylation of TAPstop in EGFR-co-expressing cells (Fig. 3A, lanes 7 and 8), although the association of EGFR and TAPstop was evident (Fig. 2C). Truncation of the p185 neu carboxyl terminus, and deletion of at least three known critical tyrosine residues, was associated with the failure to transphosphorylate the p185 neu mutant protein. Elimination of the ectodomain did not impair the intrinsic kinase activity of p185 neu -derived T⌬5, since the T⌬5 mutant receptor was still a competent tyrosine kinase (Fig. 3C, lane 1). However, unlike the full-length p185 neu , no further increase in tyrosine phos-phorylation of T⌬5 was detected in Er/T⌬5 cells with EGF stimulation (Fig. 3C, lane 2 and 3). In Er/T⌬5 cells, the EGFR was also immunoprecipitated by the anti-Bacneu antisera and still autophosphorylated after EGF treatment (Fig. 3C, lane 3). These results correlated with failure to detect physical interactions between EGFR and T⌬5 proteins (Fig. 2). Reprobing with anti-neu antibodies (Fig. 3, A and C) confirmed equivalent protein loading in paired samples with or without EGF treatment (Fig. 3, B and D). These experiments indicated that the full-length p185 neu receptor, but not mutant p185 neu proteins with NH 2 -terminal or distal COOH-terminal truncations, was able to interact with activated EGFR functionally, resulting in trans-phosphorylation.
We next analyzed tyrosine kinase activation in EGFR-positive NE91 cells with or without T691stop neu co-expression. Treatment with EGF and a chemical cross-linking reagent resulted in heavy tyrosine phosphorylation of EGFR monomers and homodimers in NE91 cells (Fig. 3E, lane 1). No detectable tyrosine phosphorylation of cytoplasmic domain-deleted T691stop neu was seen in cells with or without EGFR coexpression (Fig. 3E, lanes 2 and 3, respectively). In addition, the tyrosine phosphorylation signal of an intermediate band (ϳ285 kDa) representing EGFR⅐T691stop heterodimeric complex was also undetectable (Fig. 3E, lane 2), although a significant portion of EGFR forms a heterodimer with T691stop under these conditions (Fig. 2A, lane 5). Tyrosine kinase activation of full-length EGFR was thus completely inhibited when EGFR was physically associated with the T691stop neu mutant protein, which correlates with reduction of the transformed phenotype of primary EGFR-positive glioma cells expressing T691stop neu (42). Moreover, these results are consistent with the observation from cells co-expressing EGFR with N691stop neu derived from normal p185 neu (37).
Re-probing the membrane with an anti-EGFR antibody (CT) showed total EGFR levels in NE91 cells (Fig. 3F, lane 1), and confirmed the presence of the EGFR⅐T691stop heterodimer (ϳ285 kDa), since this complex was recognized by anti-neu in immunoprecipitation and anti-EGFR in immunoblotting (Fig.  3F, lane 2). Lysates obtained from T691stop neu-expressing cells did not react with the anti-EGFR CT probe (Fig. 3F, lane  3). Although the cytoplasmic domain deletion in T691stop did not impair heterodimerization with EGFR, the undetectable phosphotyrosine content of the intermediate heterodimer suggested that EGFR kinase activity was reduced when associated with T691stop neu. These experiments further support our model that the heteroreceptor assembly mediated primarily by ectodomain interactions facilitates kinase trans-activation and trans-phosphorylation caused by interactions between cytoplasmic domains (15,27,37).
EGF-induced Receptor Down-regulation from the Cell Surface-Numerous studies indicate that ligand-mediated receptor endocytosis and degradation is a kinase-dependent process for many types of growth factor receptors (50). We found that the efficiency of receptor down-regulation and degradation in cells co-expressing EGFR and p185 neu correlated well with heterodimeric kinase activities (37). We used this method as an alternative assay to examine the kinase activity of various heterodimers.
Cells were incubated with EGF (50 ng/ml) for various times prior to cell surface staining with anti-neu mAb 7.16.4 or anti-EGFR mAb 225 followed by the staining with fluorescein isothiocyanate-conjugated anti-mouse-IgG. Cell surface expression of each receptor was analyzed using flow cytometric analysis. EGF treatment of NE91 cells (expressing EGFR only) resulted in a reduction of cell surface EGFR, and over 60% of EGF receptors disappeared from the cell surface after 4 h of treatment (Fig. 4A). Normal EGFR down-regulation was not affected by the co-expression of T⌬5, as the efficiency of EGFR down-regulation in Er/T⌬5 cells was very similar to that seen in NE91 cells (Fig. 4A). A similar EGFR down-regulation curve was observed in Er/neu and Er/TAPstop cells (Fig. 4, C and D,  respectively), indicating that the EGFR behaves as an active receptor kinase in these cells. Moreover, about ϳ20% of p185 neu or 25% TAPstop was co-down-regulated with EGFR upon EGF stimulation (Fig. 4, C and D). As illustrated above, the low expression of p185 neu and EGFR in Er/neu cells was insufficient to demonstrate the physical association of the two receptors biochemically. The current assay was more sensitive in determining EGF-mediated receptor interactions. Control cells expressing TAPstop alone did not respond to EGF treatment, and the surface expression of TAPstop remained unchanged within the period of EGF treatment (Fig. 4D).
Analysis using an EGF-mediated pulse-chase assay showed that the down-regulated EGFR and co-down-regulated TAPstop proteins efficiently went into the degradation pathway (data not shown), similar to the cells overexpressing EGFR and p185 neu (37). Our data suggested that EGFR and either p185 neu or TAPstop associated into an active kinase complex and that these receptor assemblies exhibited comparable kinetics of receptor endocytosis.
However, co-expression of T691stop with EGFR resulted in diminished EGF-induced down-regulation of EGFR. The maximal reduction of surface EGF receptor was ϳ35% after 4 h. In addition, no detectable co-down-regulation of the cytoplasmic domain deleted T691stop was observed in Er/T691stop cells (Fig. 4B), correlating with the observation of the inactive heterodimer of EGFR⅐T691stop (Fig. 3, E and F). This finding supports the idea that receptor down-regulation is coupled to receptor tyrosine kinase activity. The formation of the inactive heterodimer between EGFR and T691stop neu proteins influenced the overall kinetics of EGFR down-regulation. Impairment of ligand-induced down-regulation of holo-EGFR by T691stop neu has also been observed in primary human cancer cells. 2 Transforming Potency of Cells Expressing Mutant p185 neu Proteins with or without EGFR-We and others have showed that the transforming potency of p185 neu requires not only its intrinsic tyrosine kinase activity (13), but also the crucial role of tyrosine phosphorylation of its carboxyl terminus, as the oncogenicity of p185 neu was greatly reduced by alteration of several tyrosine residues (41) or large structural deletions, such as seen with TAPstop (42). Transforming ability of ectodomain-deleted T⌬5 in this system was less potent than fulllength p185 neu , possibly due to the reduced efficiency of forming active receptor complexes when compared with full-length oncogenic p185 neu (27) (see Fig. 2).
We examined whether co-expression of EGFR with p185 neu and its derivatives could enhance transforming efficiency compared with cells expressing these mutant p185 neu proteins alone. Cell lines listed in Table I (except kinase-deficient  T691stop and Er/T691stop clones) were able to form foci independent of ligand stimulation (data not shown). Co-expression of EGFR with p185 neu in Er/neu cells increased the ability to form foci, both in density and absolute number (by greater than 3-fold). However, co-expression of EGFR with kinase-active truncated mutant TAPstop or T⌬5 did not enhance focus formation efficiency in Er/TAPstop and Er/T⌬5 cells when compared with TAP/stop and T⌬5 cells, respectively (data not shown).
The colony growth efficiency of these clones in soft agar is also summarized in Table I. B104-1-1 cells expressing high levels of p185 neu served as a positive control, while Er/T691stop clones served as a negative control and did not exhibit transformed colonies under the same conditions. Compared with B104-1-1 cells, cells expressing lower levels of oncogenic p185 neu formed colonies less efficiently. However, more colonies were observed in EGFR-co-expressing Er/neu cells. Co-expression of EGFR with p185 neu still permits functional heterodimerization in addition to homodimerization of either receptor, resulting in elevated biological activity, contributing to increased transforming activity in vitro. Cells expressing kinase-active truncated mutant TAPstop or T⌬5 mutant proteins alone displayed reduced colony growth efficiency in soft agar when compared with control B104-1-1 cells, although the expression levels of p185 neu variants in these cells were similar. Critically, co-expression of EGFR with T⌬5 or TAPstop did not increase colony growth efficiency in soft agar.
Tumorigenicity was studied by injection of these mutant clones individually into athymic mice. Results are presented in Table I, which summarizes receptor expression levels, tumor frequency, and tumor size. B104-1-1 cells expressing oncogenic p185 neu were used as a positive control and tumors caused by those cells appeared and grew quickly (with a latency of 5-7 days). No tumors were observed with kinase-deficient mutant clones T691stop and Er/T691stop cells (Ͼ10 weeks observation). Co-expression of EGFR and p185 neu , each at low levels, in Er/neu cells greatly accelerated tumor appearance (ϳ2 weeks), and the tumors grew aggressively when compared with p185 neu cells that also expressed low level of oncogenic p185 neu (Ͼ4 -5 weeks). Cooperative signaling between EGFR and p185 neu was thus also observed in tumorigenicity assays in vivo. T⌬5 protein expression was sufficient to induce tumors (latency period of 2-3 weeks), and TAPstop mutant receptor expression also resulted in tumor formation (latency of 4 -5 weeks). Receptor expression levels for these two mutant proteins was close to that in B104-1-1 cells. Notably, co-expression of EGFR with these mutant proteins, i.e. in Er/TAPstop and in Er/T⌬5, did not promote tumor growth.
The failure of distinct endodomain interactions between p185 neu and EGFR, caused by an ectodomain deletion (T⌬5 mutant), or the lack of a functional COOH terminus (TAPstop mutant), clearly impairs signaling needed for transformation.
EGF-dependent Cell Proliferation of Cell Lines Co-expressing EGFR and Mutant p185 neu -To analyze whether EGF-dependent heterodimerization conveys signals leading to cooperative mitogenesis, we used the MTT assay to study proliferation of various cell lines. NE91 cells expressing EGFR only served as a positive control, and showed typical EGF induction of cell growth. As expected, the maximal induction dosage of EGF was 10 ng/ml, consistent with previous observations (44). However, the maximum induction dosage of EGF in Er/neu cells was ϳ0.1 ng/ml, 2 orders of magnitude less than that observed in NE91 cells (Fig. 5A). These data suggested that p185 neu sensitized the EGF receptor responding to ligand. In contrast, the presence of T691stop in Er/T691stop cells suppressed the proliferative response to EGF, and cell growth was dramatically reduced (Fig. 5B). These results correlated with the inhibition of EGFR kinase (Figs. 3E and 4B). Interestingly, the EGFR in Er/TAPstop and Er/T⌬5 cells behaved normally in EGF-dependent mitogenesis when compared with that in NE91 cells, except that the basal growth level was higher (Fig. 5, C and D) due a more transformed phenotype (data not shown). These data correlated with previous observations , suggesting that EGFR signaling is comparable in Er/TAPstop and in Er/T⌬5 clones to that seen in NE91 cells, i.e. neither enhanced nor suppressed. However, transreceptor signaling was not observed due to either defective heterodimerization in Er/TAPstop cells or failure of heterodimerization in Er/T⌬5 cells.
EGF-dependent MAP Kinase and PI 3-Kinase Activation-To understand the mechanism underlying synergistic proliferative and transforming signal propagated by heteroreceptor interaction, we studied the EGF-induced MAP kinase and PI3 kinase pathways signaling phenomena. The proto-oncogenic p185 (Nneu) and its derivatives (N⌬5 or, N691stop) were coexpressed with EGFR, to evaluate EGF-dependent activation of downstream kinases, since p185 neu and T⌬5 are both constitutively active tyrosine kinases. An epitope-tagged HA-MAPK was also co-expressed with the combination of receptors in COS7 cells to examine downstream ERK activation.
Co-expression of p185 c-neu , but not N⌬5, with EGFR increased MAP kinase activity upon EGF stimulation. In contrast, EGFR-mediated MAP kinase activity in N691stop-coexpressing cells was suppressed when compared with cells expressing EGFR and an empty vector control (Fig. 6A). Equivalent protein expression levels of epitope-tagged HA-MAPK was also confirmed in these studies (Fig. 6B). Ectopically expressed EGFR and wild-type or mutant p185 forms were detected by immunoblot using anti-receptor specific antisera (Fig.  6, C and D). Since the intracellular domain-deleted N691stop could not be recognized by antiserum against the Nneu COOH terminus, the expression of N691stop was independently confirmed using metabolic labeled cell extracts followed by antineu immunoprecipitation (Fig. 6D, lanes 7 and 8).
Activation of PI-3-kinase requires phosphorylation of the Src homology 2-containing adapter p85 by receptor tyrosine kinases. Phosphatidylinositides are critical signaling intermediates and influence cell growth, differentiation, and adhesion (52). ErbB family members, notably ErbB-3, have been shown to associate with the p85 subunit of PI 3-kinase (53). To examine the influence of wild-type or mutant p185 on EGF-dependent activation of PI 3-kinase, plasmids expressing EGFR with vector or p185 variants were transiently expressed in COS7 cells. PI 3-kinase activity was examined in serum-starved cells with or without EGF stimulation. We observed a similar magnitude of the EGF-induced PI 3-kinase activity in cells expressing EGFR only or Er/N⌬5. The PI 3-kinase activity was much greater in Er/p185 c-neu cells, and much weaker in Er/N691stop cells (Fig. 7). Expression patterns of these receptor proteins were determined (Fig. 6, C and D).
The observed super PI 3-kinase activity in Er/p185 c-neu cells may arise through the tyrosine phosphorylation of the p85 subunit by the heteroreceptor complexes. We believe heteroreceptor complexes are more active since truncated p185 proteins alone do not seem effective at interaction with p85 (data not shown). Induced PI 3-kinase and MAPK activities therefore paralleled the heterodimerization and trans-activation events depicted in Figs. 2-4, and biological results obtained in Table I and Fig. 5. Functional heterodimerization observed in Er/neu cells permits cooperation and diversification of signaling, which contrasts with the formation of signaling-defective com-

DISCUSSION
Using p185 neu mutants, which retain the capacity to homodimerize, we observed that EGF-induced heterodimerization could occur. Heterodimerization was seen in cells co-expressing EGFR with TAPstop or T691stop mutant receptors, but not with the extracellular domain-deleted T⌬5 (Fig. 2), demonstrating that the ectodomain of p185 neu is necessary and sufficient for heterodimerization with EGFR. Indeed, heterodimerization of the EGFR and N691stop form derived from proto-oncogenic p185 neu has been observed to occur preferentially to either p185 c-neu or EGFR⅐EGFR homodimerization (37).
Two alternative assays confirmed trans-activation of ErbB family proteins following heterodimer formation. Anti-phosphotyrosine blotting showed that enhancement of tyrosine phosphorylation in response to EGF occurred only in cells co-expressing EGFR with the full-length p185 neu kinase, but not with the TAPstop or T⌬5 mutant receptors. It appears that EGFR and the T691stop neu mutant formed a kinase-inactive complex (Fig. 3), as described previously for the N691stop form (37).
An analysis of EGF-induced receptor internalization, a kinasedependent event, also indicated that receptor trans-activation is required for efficient internalization of the EGFR found in these heteromers (Fig. 4). Full-length p185 neu and TAPstop proteins were modulated by EGF and showed co-internalization with EGFR efficiently, indicating an active heterodimer was formed. Co-expression of T⌬5 with EGFR did not affect normal endocytosis of EGFR since T⌬5 could not associate with EGFR, while T691stop neu expression interfered with normal EGFR down-modulation.
Recent studies have shown that the signal adapter Grb2 is required for efficient endocytosis of EGFR (54), and selective and regulated signal transduction from activated receptor tyrosine kinases may continue within the endosome (55). Interestingly, kinase-mediated activation of ERKs may also involve endocytotic trafficking since inhibition of clathrin-mediated endocytosis has been shown to impair rapid EGF-stimulated activation of ERKs (56). Therefore, it is reasonable to speculate that EGF-induced endocytosis of these receptor complexes reflects both heterodimeric kinase activity and the efficiency of activating downstream signaling components. Indeed, fulllength p185 neu , but not other p185 neu -derived deletion mutants, displayed increased coupling of the Src homology 2-containing signaling molecule p85 to receptor activation (data not shown).
We previously reported that the co-expression of EGFR with p185 c-neu (22), but not with kinase-deficient p185 c-neu (44), synergistically transformed rodent fibroblasts. EGFR and p185 cneu associates into an active kinase complex (24) which upregulates EGF receptor function by increasing EGF binding affinity, ligand-induced DNA synthesis, and cell proliferation (23). In the current studies, when EGFR was co-expressed with oncogenic p185 neu at physiological levels (ϳ10 4 receptors/cell), we also observed enhancement of tumor growth (4-fold) in vivo and anchorage-independent growth (ϳ2-fold) in vitro, compared with the cells expressing p185 neu alone (Table I). Deletion of 122 amino acid residues from the carboxyl terminus of p185 neu eliminates three known tyrosine autophosphorylation sites (TAPstop mutant), and causes impaired cellular signaling and transforming potency (41). Overexpression of EGFR with the carboxyl-terminally truncated TAPstop mutant receptor, although leading to an active heterodimeric complex, did not recover the diminished transforming potency of TAPstop (Table I), indicating that signaling propagation through the carboxyl terminus of p185 neu could not be restored by association with full-length EGFR. These data emphasize that cooperative signaling requires not only the formation of an active kinase complex, but also a heteromeric functional carboxyl termini within the two receptor endodomains that recruit various downstream molecules required to generate signal to mediate cell growth and transformation.
The current results indicate that p185 neu ⅐EGFR heterodimerization is greatly favored, even in the presence of the neu transmembrane point mutation that facilitates p185 neu homodimerization (12). Together with the observation that ErbB-2 is the preferred heterodimerization partner of all ErbB members (36), these studies emphasize that Neu⅐ErbB-2 may mediate signaling diversity through structural interactions governed by particular ectodomain sequences. For instance, FIG. 8. The role of p185 neu subdomains in heterodimerization with EGFR and resultant signaling consequences. Functional heterodimerization requires the ectodomain for ligand-mediated physical associations, the endodomain for kinase transactivation, and the carboxyl terminus for cross-phosphorylation and combinatorial cellular signaling. Deletion of each subdomain results in inefficient heterodimerization, preventing kinase activation and defects in cooperative cellular signaling, respectively. TD5, T⌬5.
ErbB-3 is a less active kinase than other ErbB proteins (57), but serves as a binding site for Neu differentiating factor (28) and forms a potent heterodimer with ErbB-2, consequently engaging various downstream substrates. Neu⅐ErbB-2 may not be required for ligand binding, but may reconstitute signaling by laterally engaging other ErbB proteins in some preferred, but not well understood manner.
Kinase phosphorylation increases the affinity of binding of Src homology 2 and Src homology 3 domain-containing substrates, and initiates a variety of cascades. The binding of Grb2⅐Sos complexes to the active EGFR activates the Ras/Raf/ MAP kinase cascade (58). Another downstream effector whose importance in cell signaling and, potentially, in tumorigenesis is becoming increasingly understood is PI 3-kinase (52). PI 3-kinase activation has also been shown to be essential for induction of DNA synthesis by EGF (59). Current studies have demonstrated that EGF-induced ErbB heterodimers activate both the ERK and PI 3-kinase pathways. Functional wild-type heterodimers, but not defective mutant heterodimers, efficiently induce both ERK and PI 3-kinase activities, which contribute to the synergistic effects on mitogenesis and cellular transformation.
As depicted in Fig. 8, these results further support the notion that cooperative signaling caused by p185 neu ⅐EGF receptor ensembles requires the ectodomain for ligand-mediated physical association, while the intracellular domain provides contacts for efficient intermolecular kinase activation. The phosphorylated carboxyl terminus is essential for recruiting particular cellular substrates required for signal diversification.
In particular, specific ectodomain associations may therefore underlie the combinatorial interactions within the ErbB family required for signal diversification. These properties may be features that are used by many receptor ensembles involved in enzymatic signaling in cells.