c-Src-mediated Phosphorylation of the Epidermal Growth Factor Receptor on Tyr845 and Tyr1101 Is Associated with Modulation of Receptor Function*

Accumulating evidence indicates that interactions between the epidermal growth factor receptor (EGFR) and the nonreceptor tyrosine kinase c-Src may contribute to an aggressive phenotype in multiple human tumors. Previous work from our laboratory demonstrated that murine fibroblasts which overexpress both these tyrosine kinases display synergistic increases in DNA synthesis, soft agar growth, and tumor formation in nude mice, and increased phosphorylation of the receptor substrates Shc and phospholipase γ as compared with single overexpressors. These parameters correlated with the ability of c-Src and EGFR to form an EGF-dependent heterocomplex in vivo. Here we provide evidence that association between c-Src and EGFR can occur directly, as shown by receptor overlay experiments, and that it results in the appearance of two novel tyrosine phosphorylations on the receptor that are seen both in vitro and in vivo following EGF stimulation. Edman degradation analyses and co-migration of synthetic peptides with EGFR-derived tryptic phosphopeptides identify these sites as Tyr845 and Tyr1101. Tyr1101 lies within the carboxyl-terminal region of the EGFR among sites of receptor autophosphorylation, while Tyr845 resides in the catalytic domain, in a position analogous to Tyr416 of c-Src. Phosphorylation of Tyr416 and homologous residues in other tyrosine kinase receptors has been shown to be required for or to increase catalytic activity, suggesting that c-Src can influence EGFR activity by mediating phosphorylation of Tyr845. Indeed, EGF-induced phosphorylation of Tyr845 was increased in MDA468 human breast cancer cells engineered to overexpress c-Src as compared with parental MDA 468 cells. Furthermore, transient expression of a Y845F variant EGFR in murine fibroblasts resulted in an ablation of EGF-induced DNA synthesis to nonstimulated levels. Together, these data support the hypothesis that c-Src-mediated phosphorylation of EGFR Tyr845 is involved in regulation of receptor function, as well as in tumor progression.

Accumulating evidence indicates that interactions between the epidermal growth factor receptor (EGFR) and the nonreceptor tyrosine kinase c-Src may contribute to an aggressive phenotype in multiple human tumors. Previous work from our laboratory demonstrated that murine fibroblasts which overexpress both these tyrosine kinases display synergistic increases in DNA synthesis, soft agar growth, and tumor formation in nude mice, and increased phosphorylation of the receptor substrates Shc and phospholipase ␥ as compared with single overexpressors. These parameters correlated with the ability of c-Src and EGFR to form an EGF-dependent heterocomplex in vivo. Here we provide evidence that association between c-Src and EGFR can occur directly, as shown by receptor overlay experiments, and that it results in the appearance of two novel tyrosine phosphorylations on the receptor that are seen both in vitro and in vivo following EGF stimulation. Edman degradation analyses and co-migration of synthetic peptides with EGFR-derived tryptic phosphopeptides identify these sites as Tyr 845 and Tyr 1101 . Tyr 1101 lies within the carboxyl-terminal region of the EGFR among sites of receptor autophosphorylation, while Tyr 845 resides in the catalytic domain, in a position analogous to Tyr 416 of c-Src. Phosphorylation of Tyr 416 and homologous residues in other tyrosine kinase receptors has been shown to be required for or to increase catalytic activity, suggesting that c-Src can influence EGFR activity by mediating phosphorylation of Tyr 845 . Indeed, EGF-induced phosphorylation of Tyr 845 was increased in MDA468 human breast cancer cells engineered to overexpress c-Src as compared with parental MDA 468 cells. Furthermore, transient expression of a Y845F variant EGFR in murine fibroblasts resulted in an ablation of EGF-induced DNA synthesis to nonstimulated levels. Together, these data support the hypothesis that c-Srcmediated phosphorylation of EGFR Tyr 845 is involved in regulation of receptor function, as well as in tumor progression.
The epidermal growth factor receptor (EGFR) 1 is a 170-kDa single-pass transmembrane tyrosine kinase that undergoes homo-or heterodimerization and enzymatic activation following ligand binding (1,2). These events result in the trans-(auto)phosphorylation of multiple Tyr residues in the COOH-terminal tail of the molecule that serve as binding sites for cytosolic signaling proteins containing Src homology 2 (SH2) domains (3). Five sites of in vivo autophosphorylation have been identified in the EGFR: three major (Tyr 1068 , Tyr 1148 , and Tyr 1173 ) and two minor (Tyr 992 and Tyr 1086 ) (4 -7). These sites bind a variety of downstream signaling proteins which contain SH2 domains, including Shc (8) and PLC␥ (9). Binding of these or other signaling proteins to the receptor and/or their phosphorylation results in transmission of subsequent signaling events that culminate in DNA synthesis and cell division.
c-Src is a nonreceptor tyrosine kinase that functions as a co-transducer of transmembrane signals emanating from a variety of polypeptide growth factor receptors, including the EGFR (see Refs. 10 and 11, and reviewed in Ref. 12). Overexpression of wild type (wt) and dominant negative forms of c-Src in murine C3H10T1/2 fibroblasts that express normal levels of receptor, as well as experiments involving the microinjection of antibodies specific for Src family members, have revealed that c-Src is a critical component of EGF-induced mitogenesis (10,11,13). Cells which express high levels of EGFR become transformed upon continual exposure to EGF (14), and co-overexpression of c-Src in these cells dramatically potentiates their growth and malignant properties (15). Together, these findings indicate that c-Src co-operates with the EGFR in the processes of both mitogenesis and transformation.
Subsequent studies in 10T1/2 cells revealed that potentiation of EGF-induced growth and tumorigenesis by c-Src, which is observed only in cells overexpressing both c-Src and the receptor, correlates with the EGF-dependent formation of a heterocomplex containing c-Src and activated EGFR, the appearance of two unique in vitro non-autophosphorylation sites on receptors in complex with c-Src, and enhanced in vivo tyrosyl phosphorylation of the receptor substrates, PLC␥ and Shc (15). These findings suggested that c-Src-dependent phosphorylations on the EGFR can result in hyperactivation of receptor kinase activity, as measured by the enhanced ability of the receptor to phosphorylate its cognate substrates. This report identifies Tyr 845 and Tyr 1101 as c-Src-dependent sites of phosphorylation, which are present both in vitro and in vivo in receptor from 10T1/2 double overexpressing fibroblasts and from MDA468 human breast cancer cells. In the MDA468 cells, overexpression of c-Src results in a further increase in the phosphorylation of Tyr 845 , indicating that c-Src either phosphorylates this site directly or activates a secondary kinase which is responsible. Moreover, cells which transiently express EGFR bearing a Tyr to Phe mutation at Tyr 845 are impaired in their ability to synthesize DNA in response to EGF, suggesting that this c-Src mediated phosphorylation site is important for receptor function.
To create cells transiently overexpressing HER1 which contained a Tyr to Phe mutation at position 845, a DraIII-BstEII fragment from pCO11 (gift of Laura Beguinot), including the mutation at position 845, was subcloned into a pcDNA vector containing wild type human HER1 (gift of Dr. Stuart Decker, Parke Davis, Ann Arbor, MI). Neo control 10T1/2 fibroblasts were transiently transfected with 30 g of Superfect (Qiagen, Chatsworth CA) and 4 g of vector, wt HER1, or Y845F HER1 plasmid DNA according to the manufacturers' directions and incubated for 48 h.
For overexpression of c-Src in breast cancer cells, pcDNAc-Src was constructed by inserting the c-Src XhoI fragment from an existing pVZneo vector into the multicloning site of pcDNA3 (Invitrogen, San Diego, CA). MDA468 cells, obtained from N. Rosen (Sloan Kettering Cancer Center, New York), were maintained in DMEM plus 5% serum. MDA468 cells stably overexpressing chicken c-Src (clone MDA468c-Src) were generated by Lipofectin TM (Life Technologies, Inc.)-mediated gene transfer of pcDNAc-Src into parental MDA468 cells and selection with 400 g/ml G418. Parental MDA468 cells overexpress c-Src approximately 5-fold, as compared with Hs578Bst normal breast epithelial cells, and contain approximately 10 6 receptors/cell (Ref. 16), 2 while MDA468c-Src cells overexpress c-Src approximately 25-fold over levels found in normal breast epithelial cells.
Antibodies-EGFR-specific mouse monoclonal antibodies (mAbs) 3A and 4A were provided by D. McCarley and R. Schatzman of Syntex Research, Palo Alto, CA. Their derivation has been described previously and their epitopes have been mapped to residues 889 -944 and 1052-1134, respectively. EGFR-specific mAb F4, directed against amino acids 985-996, was obtained from Sigma. GD11 antibody is directed against the SH3 domain of c-Src and was characterized previously in our laboratory (17,18). Q9 antibody was raised in rabbits against the COOH-terminal peptide of c-Src (residues 522-533) and exhibits a higher affinity for c-Src than for other Src family members (19,20). Antiphosphotyrosine (Tyr(P)) antibody (4G10) was purchased from UBI (Lake Placid, NY). Negative control antibodies included pooled and purified normal rabbit or mouse immunoglobulin.
Immunoprecipitation, Western Immunoblotting, and in Vitro Kinase Assays-Methods for immunoprecipitation, Western immunoblotting, and in vitro kinase assays have been described previously (10,11,15). Cells were lysed either in CHAPS detergent buffer (10 mM CHAPS, 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 50 g/ml leupeptin, and 0.5% aprotinin), or in RIPA detergent buffer (0.25% sodium deoxycholate, 1% Nonidet P-40, 50 mM Tris-HCl, pH 7.2, 150 mM NaCl, 2 mM EDTA, 1 mM sodium orthovanadate, 50 g/ml leupeptin, and 0.5% aprotinin). Protein concentrations of detergent lysates were determined by the BCA protein assay (Pierce, Rockford, IL). 500 g of cell lysate was used for immunoprecipitations, and 50 g was used for Western blotting. For Western immunoblotting, binding of primary murine or rabbit antibodies to Immobilon membranes was detected with either 125 I-labeled goat anti-mouse IgG (NEN Life Science Products Inc.) or 125 I-protein A (NEN) used at 1 Ci/ml, specific activity 100 Ci/ml. For kinase assays, immunoprecipitates were prepared in and washed twice in CHAPS buffer, then washed twice with HBS buffer (150 mM NaCl, 20 mM HEPES, pH 7.4). Each kinase reaction was conducted in 20-l volumes containing 20 mM PIPES, pH 7.5, 10 mM MnCl 2 , and 10 Ci of [␥-32 P]ATP (6000 Ci/mmol, NEN) for 10 min at room temperature. Incubations were terminated by addition of sample buffer, and labeled products were resolved by SDS-PAGE and visualized by autoradiography.
In Vitro Binding and Far Western Analysis Using GSTc-SrcSH2-The construction and preparation of GST fusion proteins containing the SH2 domain of c-Src was described previously (21). To reconstitute binding between tyrosyl-phosphorylated EGFR and the SH2 domain of c-Src, 2 g of immobilized GST-c-SrcSH2 fusion protein was incubated with 100 g of 10T1/2 cell lysate protein prepared in RIPA buffer. After 3 h gentle mixing at 4°C, beads were washed three times with RIPA buffer, resuspended in SDS sample buffer, and boiled. Eluted proteins were separated by SDS-PAGE, transferred to Immobilon, and immunoblotted with either the Tyr(P) or the 3A/4A or F4 (Sigma) monoclonal EGFR antibodies.
To assess direct binding of GST-c-SrcSH2 to the EGFR, receptor from 500 g of cell lysate protein in RIPA buffer was immunoprecipitated with 3A/4A mAbs. The resulting EGFR immunoprecipitates were resolved by SDS-PAGE, transferred to Immobilon membranes, and incubated with 1 mg/ml purified GST-c-SrcSH2 fusion protein in blocking buffer at 4°C overnight. The membrane was then probed with 1 g/ml affinity purified, polyclonal rabbit anti-GST antiserum in blocking buffer, 3 and immunoglobulin binding was detected by 125 I-protein A.
Metabolic Labeling-NeoR1 or 5HR11 cells were grown to 50 -75% confluency in 150-mm dishes, washed with phosphate-free DMEM, and incubated for 18 h in phosphate-free DMEM containing 0.1% dialyzed fetal bovine serum and 1 mCi/ml [ 32 P]orthophosphate (NEN Life Science Products Inc.) in a final volume of 10 ml. For pervanadate treatment, labeling medium was adjusted to a concentration of 3 mM H 2 O 2 and 5 M Na 3 VO 4 just prior to EGF stimulation. Cells were stimulated in the presence of pervanadate by addition of 100 ng/ml EGF to the labeling medium for 5 min, washed twice with phosphate-free DMEM, and lysed in CHAPS detergent buffer. Extract from an entire plate (approximately 1-2 mg of protein) was immunoprecipitated with c-Src or EGFR-specific antibodies as described above.
Two-dimensional Tryptic Phosphopeptide Analysis-Immunoprecipitates of in vitro or in vivo 32 P-labeled EGFR were resolved by SDS-PAGE. The EGFR was localized by autoradiography, excised from the gel, and digested with trypsin as described by Boyle et al. (22). Phosphotryptic peptides were separated by electrophoresis at pH 1.9 in the first dimension and ascending chromatography in the second dimension on cellulose thin layer chromatography (TLC) plates. Chromatography buffer contained isobutyric acid, 1-butanol, pyridine, acetic acid, H 2 O (125:3.8:9.6:5.8:55.8). Migration of synthetic phosphopeptides was detected by spraying the dried TLC plate with a hypochlorite solution consisting of sequential sprays with 10% commercial Clorox, 95% ethanol, 1% potassium iodide, and saturated o-tolidine in 1.5 M acetic acid, as described in Stewart and Young (23).
High Performance Liquid Chromatography (HPLC)-For HPLC analysis of peptides derived from the EGFR associated with c-Src, 32 P-labeled phosphotryptic peptides were prepared as above and suspended in 0.05% trifluoroacetic acid. Peptides were injected into a Perkin-Elmer Series 4 Liquid Chromatograph equipped with a Vydac C18 column (4.6 ϫ 250 mm) and eluted with increasing concentrations of acetonitrile (0 to 100%) at a flow rate of 1 ml/min, as described by Wasilenko et al. (24). 500-l fractions were collected, and Cerenkov counts of each fraction were determined. Fractions containing peptides "0" and "3" were identified by two-dimensional TLC analysis for their ability to co-migrate with the appropriate peptide in a mixture of total in vitro phosphorylated receptor peptides. Appropriate fractions were then lyophilized and subjected to Edman degradation.
Edman Degradation-HPLC fractions of 32 P-labeled EGFR phosphotryptic peptides or spots eluted from TLC plates were subjected to automated Edman degradation, as performed by the University of Virginia Biomolecular Research Facility. Briefly, phosphorylated peptides were coupled to a Sequelon aryl amine membrane (25), washed with 4 ϫ 1 ml of 27% acetonitrile, 9% trifluoroacetic acid, and 2 ϫ 1 ml of 50% methanol, and transferred to an applied Biosystems 470A sequenator using the cartridge inverted as suggested by Stokoe et al. (26). The cycle used for sequencing was based on that of Meyer et al. (27), but modified by direct collection of anilinothiazolinone amino acids in neet trifluoroacetic acid as described by Russo et al. (28). Radioactivity was measured by Cerenkov counting.
Identification of Peptides 0 and 3-Phosphorylated peptides (corresponding to residues GMN(Y-P)LEDR, candidate for peptide 3; or E(Y-P)HAEGGK, candidate for peptide 0) were synthesized by the University of Virginia Biomolecular Research Facility. Synthetic peptides were mixed with oxidized in vitro labeled phosphotryptic peptides from c-Src-associated EGFR, separated on cellulose TLC plates, and visualized by spraying with the hypochlorite solution as described above. One candidate for peptide 3 (GMNYLEDR) was synthesized as a phosphopeptide and tested for comigration as above. Another candidate for peptide 3 (DPHY 1101 QDPHSTAVGNPEYLNTVQPTCVNSTF DSPAH-WAQK), which was too large to chemically synthesize, was tested by further digestion of in vitro labeled peptide 3 with a proline-directed protease (Seikagaku, Rockville, MD), according to the method of Boyle et al. (22). In brief, the spot corresponding to peptide 3 was scraped off the TLC plate, eluted with pH 1.9 buffer, and digested with 5 units of proline-directed protease in 50 mM ammonium bicarbonate at pH 7.6 at 37°C for 1 h. Peptides were separated by two-dimensional electrophoresis as described above.
BrdUrd Incorporation-Neo control cells, which had been transfected with cDNAs encoding wild type EGFR, Y845F EGFR, or vector alone were cultured for 48 h, then serum starved for an additional 30 h prior to the administration of 100 M BrdUrd and either 100 ng/ml EGF or 10% fetal calf serum in fresh growth medium. Treated cells were incubated for 18 h and co-stained for HER1 expression and BrdUrd incorporation as described by the manufacturer of the BrdUrd-specific antibody (Boehringer Mannheim). Briefly, fixed cells were treated with 2 N HCl for 1 h at 37°C and incubated with a mixture of primary antibodies (1:100 dilution of the HER1-specific Ab-4, and a 1:15 dilution of anti-BrdUrd mouse antibody in serum-free medium for 1 h at 37°C), followed by incubation with a mixture of secondary antibodies (75 g/ml fluorescein isothiocyanate-conjugated goat anti-rabbit IgG and 4 g/ml Texas Red-conjugated goat anti-mouse IgG) for 1 h at 37°C. Both secondary reagents were obtained from Jackson Immunoresearch Laboratories, West Grove, PA.

Direct Binding of c-Src SH2 Domain to the EGFR-Previous
work from our laboratory demonstrated a synergistic interaction between c-Src and the EGFR which led to increased cell growth and tumor development (10,11,15). This functional synergism was most striking when cells overexpressed both c-Src and the EGFR (5HR cells) and correlated with the ability of c-Src and the EGFR to form specific, EGF-dependent heterocomplexes in vivo. The formation of this c-Src⅐EGFR complex raises the question of whether binding between c-Src and the EGFR occurs directly, or is mediated by another protein present in the complex. To test whether association could be mediated by a Tyr(P)-SH2 interaction, lysates from unstimulated and stimulated Neo, 5H, NeoR, or 5HR cells were incubated with a GST-c-SrcSH2 bacterial fusion protein linked to agarose beads, and precipitated proteins were probed with Tyr(P) antibody. Fig. 1, panel A, lanes 4 and 8, show that a tyrosylphosphorylated protein of 170 kDa was precipitated by GST-c-SrcSH2 from extracts of cells overexpressing the EGFR after activation of the receptor with EGF. This 170-kDa protein co-migrated with the EGFR precipitated with receptor-specific mAbs 3A/4A (data not shown). Other proteins that bound c-SrcSH2 included p125 FAK (21), which was detected in all the cell lysates, a 75-80-kDa protein, cortactin, which was most prominent in 5H cells (30), and a 62-kDa protein, presumed to be related to the 62-kDa "DOK" protein associated with p120 Ras-GAP (31)(32)(33)(34). These results suggest that in vivo, multiple Tyr(P)-containing proteins in addition to the EGFR are capable of interacting with c-Src via its SH2 domain and contribute to the highly tumorigenic phenotype of the double overexpressing cells. Incubation of cell extracts with GST-beads alone resulted in no detectable binding of Tyr(P)-containing proteins (data not shown).
To confirm that the 170-kDa protein was indeed the EGFR, lysates prepared from unstimulated and stimulated NeoR and 5HR cells were precipitated with immobilized GST-c-SrcSH2, and bound proteins were immunoblotted with EGFR-specific mAbs 3A/4A. Fig. 1, Panel B, demonstrates that receptor antibody detected the 170-kDa protein only in stimulated cells, as in Panel A, confirming its identity as the EGFR. To test if the interaction between the activated EGFR and c-SrcSH2 could be direct, receptor immunoprecipitates were subjected to a "Far Western" overlay experiment, using GST-c-SrcSH2, GST-specific antibody, and 125 I-protein A. Fig. 1, Panel C, lanes 2 and 4, shows that GST-c-SrcSH2 bound the EGFR and, as predicted, the interaction required activation by EGF. GST alone exhibited no binding (data not shown). Panel D verified that nearly equal amounts of receptor were present in all immunoprecipitates. These results provide evidence for the involvement of SH2-Tyr(P) interactions in the formation of the EGFR⅐c-Src complex.
In Vivo and in Vitro Phosphorylation of Novel, Non-auto- In vitro association between activated EGFR and c-Src SH2 domain. 500 g of lysate protein from the indicated nonstimulated cells or cells stimulated with 100 ng/ml EGF for 2 min was incubated with GST-c-SrcSH2 fusion protein immobilized on glutathione-agarose beads (Panels A and B) or EGFR mAbs 3A/4A bound to Protein A (Panels C and D), as described under "Materials and Methods." Affinity-precipitated proteins were washed and subjected to SDS-PAGE, transferred to Immobilon membranes, and probed with: A, Tyr(P) mAb 4G10; B, EGFR mAbs 3A/4A; C, GST-c-Src SH2 fusion protein; and D, EGFR mAb 3A/4A. Binding of primary antibody was visualized by incubating membranes with 125 I-labeled goat anti-mouse IgG (Panels A, B, and D), and binding of GST-c-SrcSH2 fusion protein was detected by rabbit anti-GST and 125 I-protein A (Panel C). GST-c-SrcSH2 fusion protein is shown to bind directly to activated EGFR.
phosphorylation Sites on the EGFR in Complex with c-Src-Overexpression of both EGFR and c-Src in 10T1/2 cells results in increased tyrosyl phosphorylation of receptor substrates, PLC␥ and Shc, following EGF treatment (15). These findings suggest that the c-Src-associated receptor is modified in some manner as to increase its kinase activity. To examine the receptor for novel phosphorylations, the in vitro phosphorylated, c-Src-associated 170-kDa protein was excised from the gel and subjected to two-dimensional phosphotryptic peptide analysis. The phosphopeptide map of c-Src-associated receptor was then compared with the map of the free receptor, immunoprecipitated with receptor antibody. Fig. 2, Panels A and B, demonstrate that the maps are nearly identical; however, two additional phosphorylations (designated peptides 0 and 3) were seen in the map of the EGFR complexed with c-Src, suggesting that c-Src was responsible for their phosphorylation. Consistent with this notion, two-dimensional phosphoamino acid analysis of the in vitro labeled EGFR demonstrated that peptides 0 and 3 contained only phosphotyrosine (data not shown). Panel C shows that the two novel phosphopeptides were also detected in the receptor found in complex with c-Src from 32 P metabolically labeled 5HR cells that had been treated with pervanadate and EGF for 5 min. These data indicate that two phosphorylations occur on the EGFR both in vitro and in vivo when c-Src becomes physically associated with the receptor following EGF stimulation.
Initial attempts to detect peptides 0 and 3 in receptor immunoprecipitations from 32 P-labeled NeoR or 5HR cells yielded phosphopeptide maps that contained peptide 3 but no or barely detectable levels of peptide 0 (Fig. 3, Panels A and C). Neither could peptide 3 nor peptide 0 be detected reproducibly in receptor that was associated with c-Src from 5HR cells (data not shown). Furthermore, in receptor immunoprecipitations, the levels of peptide 3 derived from NeoR versus 5HR cells appeared nearly equal (compare Panels A and C), suggesting that peptide 3 may not be an in vivo, c-Src-dependent site of phosphorylation. In these experiments, lysates were prepared in CHAPS buffer containing a mixture of conventional protease and phosphatase inhibitors, including orthovanadate (see "Materials and Methods"). However, modification of the EGF treatment regimen to include pervanadate during stimulation allowed us to detect peptide 0 in receptor immunoprecipitates from NeoR (Panel B) and 5HR (Panel D) cells. These conditions revealed more peptide 0 in receptor from 5HR than from NeoR cells, confirming the ability of c-Src to modulate the phosphorylation of this peptide. Of special note was the finding that peptide 0 was the only peptide seen to increase in phosphoryl-ation in response to pervanadate treatment, suggesting that its phosphorylation is more labile than that of peptide 3 or the other phosphorylations on the receptor, which presumably correspond to autophosphorylation sites. Together with the in vitro studies depicted in Fig. 2, the results from the in vivo experiments indicate that peptide 0 is an in vitro and in vivo site of receptor phosphorylation that is regulatable by c-Src. Following this line of reasoning, the low level of peptide 0 phosphorylation seen in receptor immunoprecipitates from NeoR cells (Fig. 3, Panel B) could be due to endogenous c-Src. However, the involvement of other tyrosine kinases in the in vivo phosphorylation of peptide 0 cannot be ruled out.
Whether c-Src alone plays a role in regulating the phospho- B and D) or not (Panels A and C) along with 100 ng/ml EGF for 5 min prior to lysis in RIPA detergent buffer. EGFR was immunoprecipitated with mAbs 3A/4A, and the receptor was processed for phosphotryptic analysis as described in the legend to Fig. 3. Panel A, EGFR from NeoR cells; Panel B, EGFR from pervanadate-treated NeoR cells; Panel C, EGFR from 5HR cells; Panel D, EGFR from pervanadate-treated 5HR cells. ϳ3000 cpm were loaded per TLC plate. TLC plates were exposed to Pegasus blue film for 18 h .   FIG. 2. EGFR phosphotryptic peptides radiolabeled in vitro or in vivo. For in vitro labeling (Panels A and B), 5HR and NeoR cells were stimulated with 100 ng/ml EGF for 30 min, followed by lysis in CHAPS buffer and immunoprecipitation of extract proteins with either c-Src-specific (GD11) or EGFR-specific (3A/4A) antibody. Precipitated proteins were then subjected to an in vitro kinase reaction, and products were analyzed by SDS-PAGE and autoradiography. For in vivo experiments (Panel C), cells were labeled for 18 h in phosphate-free media containing [ 32 P]orthophosphate, stimulated with 100 ng/ml EGF for 5 min in the presence of pervanadate, and lysed in CHAPS buffer. Extracts were immunoprecipitated with GD11 antibody, and precipitated proteins were analyzed by SDS-PAGE and autoradiography. c-Src-associated, 32 Plabeled EGFR was eluted from gel slices, and samples were trypsinized and analyzed by two-dimensional TLC as described previously (17). Labeled peptides were visualized by autoradiography. Panel A, in vitro labeled EGFR immunocomplexes from NeoR cells (2000 cpm); Panel B, in vitro labeled c-Src-associated EGFR from 5HR cells (2000 cpm); Panel C, c-Src-associated EGFR from 5HR cells labeled in vivo (3000 cpm). Tryptic maps were exposed to Pegasus Blue film (Pegasus, Burtonsville, MD) for 18 h. rylation of peptide 3 in vivo is less clear. In vitro, peptide 3 phosphorylation appears to be unique to the receptor associated with c-Src (compare Panels A and B of Fig. 2), and HPLC analysis corroborates this, where phosphorylation of the peak corresponding to peptide 3 was found to be ϳ3.5-fold greater when the receptor was associated with c-Src versus free receptor (data not shown). Furthermore, the level of in vivo phosphorylation of peptide 3 in the c-Src-associated receptor is greater than that found in the "free" receptor (compare Fig. 2, Panel C, with Fig. 3, Panel D). However, peptide 3 is readily detected in free receptor labeled in vivo, and its level of phosphorylation does not appear to increase to any great extent in 5HR versus NeoR cells (Fig. 3, Panels B and D). These data can be interpreted to mean either that peptide 3 contains a nonlabile site of phosphorylation, regulatable by c-Src (in contrast to peptide 0), or that phosphorylation of peptide 3 may be regulated by an additional tyrosine kinase in vivo.

FIG. 3. Phosphorylation of peptides 0 and 3 in metabolically labeled pervanadate-treated cells. NeoR and 5HR cells were incubated for 18 h with [ 32 P]orthophosphate as above. Pervanadate (3 mM H 2 O 2 and 5 M Na 3 VO 4 ) was added (Panels
To identify the amino acids phosphorylated in vitro in a c-Src-dependent manner, fractions containing peptides 0 and 3 were isolated by HPLC. Peptide 0 eluted at 8.5% acetonitrile, while peptide 3 eluted at 10.5% acetonitrile (not shown). These HPLC fractions, which were of greater than 95% purity, were subjected to sequential Edman degradation to determine the cycle number at which radioactivity was released. Results from these analyses indicated that a phosphoamino acid residue was located at the second position of peptide 0 (Fig. 4, Panel A) and at the fourth position of peptide 3 (Fig. 4, Panel B). Of the tryptic peptides generated from the intracellular domain of the EGFR which contain Tyr residues, those peptides containing Tyr 845 , Tyr 867 , or Tyr 891 were potential candidates for peptide 0, while those peptides containing Tyr 803 or Tyr 1101 were potential candidates for peptide 3 (see Table I). The Tyr 845 -containing peptide was selected for further study as a candidate for peptide 0, since it showed 50% homology to sequences contained within the autophosphorylation site of Src (Tyr 416 ), indicating that it could be a potential c-Src target. The octamer composed of E(P-Y 845 )HAEGGK (peptide 0) was chemically synthesized to include a phosphorylated Tyr 845 and analyzed either alone (Fig. 5, Panel A) or in a mixture with total peptides from in vitro labeled, c-Src-associated receptor by two-dimensional TLC (Panel C). The synthetic octamer comigrated with peptide 0 in the mixture, thereby identifying Tyr 845 as the phosphorylated residue in peptide 0.
Since peptides 0 and 3 migrated similarly in the two-dimensional chromatography, it was expected that they would share similar isoelectric points and hydrophobicities. Both candidates for peptide 3 (GMNY 803 LEDR or DPHY 1101 QDPHSTAVGNPEYLNTVQPTCVNSTFDSPAHWAQK, see Table I) had theoretical isoelectric points and calculated hydrophobic indices (22) similar to those of the Tyr 845 -containing peptide, indicating that both were potential candidates. The Tyr 803 -containing peptide was selected first for further study, since it was smaller and more easily synthesized. However, this synthetic phosphopeptide did not co-migrate with peptide 3 nor with any of the other EGFR phosphopeptides (data not shown), indicating that the Tyr 1101 -containing peptide was the preferred candidate. To verify the identity of peptide 3, in vitro labeled peptide 3 was scraped off the TLC plate, eluted with pH 1.9 buffer, and subjected to further digestion with a prolinedirected protease as described under "Materials and Methods." Since, of the two candidate peptides, only the Tyr 1101containing peptide contains proline residues, any change in mobility resulting from digestion with this protease would confirm its identity as peptide 3. As a control, peptide 0, which does not contain any proline residues, was digested with proline-directed protease and no change in mobility was observed (data not shown). Fig. 6 shows that digestion of spot 3 with the proline-directed protease resulted in a change of migration primarily in the first dimension (compare Panel A with Panel B). To confirm that a mobility shift was indeed occurring, digested and undigested peptide 3 were mixed (Panel C). The results identify peptide 3 as Tyr 1101 .
Phosphorylation of Tyr 845 and Tyr 1101 in HER1 from Breast Tumor Cells-Our laboratory has previously demonstrated the presence of EGF-dependent c-Src⅐EGFR heterocomplexes in several human breast tumor cell lines including MDA468, which overexpresses both c-Src and HER1 (16). Since the presence of this heterocomplex is correlated with general increases in downstream receptor-mediated signaling and tumorigenicity in these cells, as compared with cell lines which do not overexpress the EGFR, we wished to investigate whether Tyr 845 and/or Tyr 1101 were phosphorylated in c-Src-associated EGFR derived from breast tumor cells. Fig. 7 demonstrates that phosphopeptides 0 and 3 are both present in in vitro labeled, c-Src-associated EGFR from EGF-stimulated MDA468 cells, although peptide 0 is weakly detected in the absence of pervanadate treatment. To further investigate the role of c-Src in mediating the phosphorylation of these sites, an MDA468 derivative cell line which stably overexpresses c-Src approximately 25-fold over levels in normal breast epithelial cells (MDA468c-Src cells, Panel B) was created. In these cells, the phosphorylation of peptide 0 (Tyr 845 ) was greatly enhanced, while the phosphorylation of peptide 3 (Tyr 1101 ) was unchanged (Panel C).
Role of Tyr 845 in EGF-dependent Mitogenesis-A tyrosyl residue homologous to Tyr 845 is conserved in many other receptor tyrosine kinases, and mutation of these conserved tyrosines to phenylalanine results in a reduced ability of the receptors to signal downstream events (35)(36)(37). Thus, it is possible that mutation of Tyr 845 to phenylalanine would likewise decrease EGF-dependent signaling through the EGFR. To directly test the requirement of Tyr 845 phosphorylation for receptor function, a variant receptor bearing a Y845F mutation was transiently transfected into Neo cells, and the effects on DNA synthesis were assayed by measuring bromodeoxyuridine (BrdUrd) incorporation in response to EGF (Fig. 8). The level of BrdUrd incorporation in cells expressing the Y845F mutant EGFR was reduced to approximately 30% of that induced by the wild type receptor, indicating that the mutant EGFR could interfere with the function of endogenous receptor and was thus acting in a dominant negative manner. Similar results were obtained when Y845F receptor was expressed in cells which overexpress c-Src (38). These findings suggest that phosphorylation of Tyr 845 is necessary for the mitogenic function of the receptor.

DISCUSSION
Previous studies from our laboratory using the C3H10T1/2 murine fibroblast model demonstrated that simultaneous overexpression of c-Src and EGFR potentiates EGF-dependent mitogenesis, transformation, and tumorigenesis, as well as EGFdependent association of c-Src with the receptor and increases in tyrosyl phosphorylation of the receptor substrates Shc and PLC␥ (15). These events correlated with the appearance of two novel tyrosine phosphorylation sites on the receptor, suggesting that one mechanism by which c-Src could synergize with the EGFR is by physically complexing with it and mediating the phosphorylation of novel non-autophosphorylation tyrosine residues, which in turn may result in hyperactivation of the receptor and enhanced phosphorylation of receptor substrates. This increased signaling would then culminate in augmented cell division and tumor growth. Such a model was recapitulated in breast cancer cell lines of epithelial origin, wherein cell lines that express high levels of c-Src and EGFR exhibit EGF-de- Glu-Tyr-His-Ala-Glu-Gly-Gly-Lys 867 Ile-Tyr-Thr-His-Gln-Ser-Asp-Val-Trp-Ser-Tyr-Gly-Val-Thr-Val Trp-Glu-Leu-Met-Thr-Phe-Gly-Ser-Lys 891 Pro-Tyr-Asp-Gly-Ile-Pro-Ala-Ser-Glu-Ile-Ser-Ser-Ile-Leu-Glu Lys 803 Peptide 3 Gly-Met-Asn-Tyr-Leu-Glu-Asp-Arg 1101 Asp-Pro-His-Tyr-Gln-Asp-Pro-His-Ser-Thr-Ala-Val-Gly-Asn Pro-Glu-Tyr-Leu-Asn-Thr-Val-Gln-Pro-Thr-Cys-Val-Asn-Ser Thr-Phe-Asp-Ser-Pro-Ala-His-Trp-Ala-Gln-Lys pendent association between c-Src and the receptor, augmented signaling through Shc and MAP kinase, and enhanced tumor formation, as compared with breast tumor cell lines which do not overexpress both c-Src and the EGFR (16). Because these and other studies link c-Src and the EGFR etiologically to tumorigenesis and malignant progression in many human tumors (reviewed in Ref. 12), identification of the two novel c-Src-dependent phosphorylations on the receptor and determination of their functions has taken on added importance, as they represent possible sites for therapeutic intervention.
Here we identify these c-Src dependent sites as Tyr 845 and Tyr 1101 and demonstrate that they become phosphorylated in murine fibroblasts both in vitro and in vivo in c-Src/EGFR double overexpressing cells in an EGF-dependent manner. Enhanced phosphorylation of Tyr 845 was also observed in MDA468 human breast cancer cells when c-Src was overexpressed, indicating that such phosphorylations can occur in cells of both mesodermal and epithelial origin. More importantly, the fact that cells expressing a Y845F variant of the EGFR are impaired in their ability to synthesize DNA in response to EGF treatment provides direct evidence for the importance of this phosphorylation. Together, these findings support the hypothesis that the c-Src-mediated phosphorylation of Tyr 845 is a critical event for EGFR function, and in certain situ-ations where overexpression of these molecules exists (such as in certain breast tumors), the increased receptor signaling resulting from this phosphorylation could lead to enhanced tumorigenesis.
Tyr 845 resides in an intriguing position on the receptor, namely in the activation lip of the kinase domain (39,40). Amino acid sequences in this lip are highly conserved among tyrosine kinases (41). Crystallographic studies indicate that phosphorylation of Tyr 845 homologues stabilizes the activation lip, maintains the enzyme in an active state, and provides a binding surface for substrate proteins; while mutation of these sites in their respective receptors results in decreases in cell growth and transformation (37, 40 -43). A similar situation appears to exist for the EGFR, as cells expressing the Y845F variant receptor showed decreases in their ability to respond mitogenically to EGF. This impairment of DNA synthesis occurred both in a background of endogenous levels of c-Src, as shown here, as well as in cells where c-Src was overexpressed (38). This finding argues that endogenous levels of c-Src are capable of mediating the phosphorylation of Tyr 845 and that the Y845F form of the receptor acts in a dominant negative fashion. Which downstream targets of the receptor are affected in various cell types by the Y845F mutation is not known. Other studies from our laboratory demonstrate that EGF-induced increases in Shc and mitogen-activated protein kinase tyrosyl The octapeptide E(Y-P)HAEGGK was synthesized to contain phosphorylated Tyr 845 and analyzed by two-dimensional electrophoresis/chromatography on TLC plates, either alone (Panel A) or in a mixture with total in vitro labeled tryptic phosphopeptides derived from the receptor which co-precipitated with c-Src (Panel C). The synthetic phosphopeptide, detected by hypochlorite spraying, comigrated with tryptic peptide 0, verifying Tyr 845 as the site on the receptor whose phosphorylation is dependent on c-Src. Panel B, total phosphopeptides from c-Src-associated receptor alone. Panel D, sequence homology between the peptide containing Tyr 416 of c-Src and the peptide containing Tyr 845 of the EGFR. 3000 cpm of in vitro labeled tryptic phosphopeptides were loaded along with 2 g of synthetic phosphopeptide.
FIG. 6. Identification of peptide 3. In vitro phosphorylated peptide 3 (as in Fig. 5B) was scraped and eluted from the TLC plate and subjected to digestion with proline-directed protease. Undigested or digested, eluted peptide 3 was then analyzed by two-dimensional TLC either alone (Panels A and B, respectively) or mixed (Panel C). The altered mobility of digested peptide 3 indicates the presence of a proline in the sequence and identifies the peptide as containing Tyr 1101 . 100 cpm of either digested or undigested peptide 3 were loaded on each TLC plate. phosphorylation occur normally when the Y845F receptor is transiently co-expressed in COS cells (38). This finding suggests that a mitogen-activated protein kinase-independent pathway plays a more dominant role in mitogenic signaling emanating from the receptor when it is phosphorylated on Tyr 845 .
That phosphorylation of this Tyr 845 residue may regulate receptor activity is consistent with the observation that a Tyr 845 homologue is not found in the EGFR family member erbB3/HER3, which is known to lack kinase activity (44). However, unlike the situation resulting from mutation of the analogous site in other receptor tyrosine kinases, mutation of Tyr 845 does not appear to alter the EGF receptor's ability to autophosphorylate or to phosphorylate the downstream substrate, Shc (38). In many tyrosine kinases, including Src, JAK 2, and receptors for colony stimulating factor-1, platelet-derived growth factor, insulin, and fibroblast growth factor, the Tyr 845 homologue is an autophosphorylated residue (35,36,(45)(46)(47)(48). However, to date Tyr 845 has not been identified as an autophosphorylation site for the EGF receptor. This could be due to the highly labile nature of the phosphorylation and/or to the fact that c-Src appears to regulate its phosphorylation (see Figs. 2, 3, and 7). Together these findings raise a number of questions: namely, whether c-Src phosphorylates Tyr 845 directly, whether binding of c-Src to the receptor causes the receptor to phosphorylate itself, or whether another tyrosine kinase which mediates the phosphorylation is recruited into the complex or activated by c-Src.
Several pieces of evidence support the hypothesis that c-Src phosphorylates the receptor directly. First, Tyr 845 is homologous to Tyr 416 in Src, which is an autophosphorylation site for Src (39). Additional evidence comes from our studies with both 10T1/2 murine fibroblasts and MDA468 breast cancer cells overexpressing c-Src, where an enhanced phosphorylation of Tyr 845 is observed. Moreover, other studies from our laboratory demonstrate that overexpression of a kinase inactive form of c-Src in 10T1/2 cells or in MDA468 cells results in a striking decrease in Tyr 845 phosphorylation (38). 4 These latter findings indicate that c-Src kinase activity is necessary for the phosphorylation of Tyr 845 and strongly argue that Tyr 845 is a direct substrate of c-Src. Last, in vitro affinity precipitation and Far Western analyses (Fig. 1, this report, and Refs. 29, 49, and 50) demonstrate that the c-Src SH2 domain can bind activated EGFR specifically and directly, suggesting that recruitment of other tyrosine kinases is not necessary to mediate the phosphorylation of Tyr 845 . However, other EGFR family members (including HER2/neu) (2,51,52) and several cytosolic tyrosine kinases, such as other c-Src family members (13) and JAK kinases (53,54), have been reported to be involved in receptormediated signaling, and we cannot exclude their possible involvement in phosphorylation of Tyr 845 or of Tyr 1101 . Whether simple binding of c-Src induces a conformational change in the receptor so that it can autophosphorylate is a much more difficult question to address, a question that minimally awaits identification of the c-Src-binding site.
Other investigators have also described Src-mediated phosphorylations on the EGFR, and Wasilenko et al. MDA468 or MDA468c-Src cells were stimulated with 100 ng/ml EGF for 30 min, followed by lysis in CHAPS buffer and immunoprecipitation of extract proteins with either c-Src-specific (GD11) or EGFR-specific (F4) antibody. Precipitated proteins were then subjected to an in vitro kinase reaction. The labeled EGFR was eluted from gel slices, and samples were trypsinized and processed as described previously in the legend to Fig. 3. Labeled peptides were visualized by autoradiography. Panel A, phosphotryptic peptides from in vitro labeled EGFR immunocomplexes from MDA468 cells (4000 cpm). Panel B, protein extracts (50 g) from MDA468 parental, 5HR, or MDA468c-Src cells which overexpress c-Src, were separated by SDS-PAGE and subjected to immunoblotting with GD11 antibody. Panel C, phosphotryptic peptides from in vitro labeled, c-Src-associated EGFR from MDA468c-Src cells (4000 cpm).
FIG. 8. Phosphorylation of Tyr 845 is required for EGF-induced DNA synthesis. Neo control cells were transfected with plasmid DNA encoding Y845F or wild type EGFR, cultured for 2 days, serum starved for 30 h, and left untreated or treated with 40 ng/ml EGF for an additional 18 h. Cells were fixed and co-stained for EGFR expression and BrdUrd incorporation. Results are expressed as the mean percent Ϯ S.E. of cells expressing EGFR that were positive for BrdUrd incorporation. Thirty-five to 75 cells were analyzed for each variable in three independent experiments. strated that in NIH3T3 cells co-expressing the transforming oncoprotein v-Src along with EGFR, the receptor contained several novel sites of tyrosine phosphorylation, one of which they postulated might be Tyr 845 (SPY1). Sato et al. (55) provide additional evidence for phosphorylation of Tyr 845 in A431 cells in a c-Src-dependent fashion, while Stover et al. (56) showed that Tyr 891 and Tyr 920 were phosphorylated in the c-Src-associated EGFR derived from MCF7 cells. However, neither we nor Sato et al. (55) have been able to detect phosphorylation of Tyr 891 or Tyr 920 , and none of these reports have linked the various phosphorylations to biological changes in receptor activity (e.g. mitogenesis, tumorigenesis). Thus, while there is some discrepancy among the different cell systems, our data and those of others indicate that Tyr 845 is a major c-Src-dependent phosphorylation site on the EGFR, and that it is associated with increases in receptor function. These findings suggest that multiple tyrosine phopshorylations may be regulated by c-Src.
A potential role for Tyr 1101 is more unclear, as this residue is not conserved among EGFR family members and its phosphorylation level in vivo is not as noticeably altered upon c-Src overexpression as is that of Tyr 845 (see Fig. 3). However, Tyr 1101 may function as a docking site for novel or known signaling proteins, perhaps in an SH2-dependent manner similar to that of the other autophosphorylation sites in the COOH terminus. One of the candidate binding proteins is c-Src itself. In peptide inhibition experiments using synthetic peptides to inhibit the binding between the EGFR and the SH2 domain of c-Src, the SH2 domain of c-Src was shown to bind Tyr 992 (49,55) and Tyr 1101 (50) preferentially. Thus, c-Src could bind one of these sites, which could position it to phosphorylate Tyr 845 . In MDA468 breast cancer cells, Tyr 845 appeared to be the site most affected by c-Src. While the data from the 10T1/2 system suggests that the phosphorylation of both Tyr 1101 and Tyr 845 is dependent on c-Src, it may be that the phosphorylation of each peptide turns over at different rates in different cell types. Also, the endogenous levels of c-Src in the parental MDA468 cells may be capable of phosphorylating Tyr 1101 to a maximal extent, and no further phosphorylation could result from overexpression. In this regard, overexpression of c-Src may allow for maximal phosphorylation of Tyr 845 if this phosphorylation turns over at a faster rate, which appears to be the case as the results from Fig. 4 indicate.
Our data show that phosphorylation on Tyr 845 appears to be critical for EGFR-mediated mitogenesis. Moreover, our results ( Figs. 3 and 7) suggest that basal levels of c-Src are able to mediate phosphorylation of Tyr 845 to some extent, and that this phosphorylation is important to receptor function. In a cell where overexpression and/or activation of c-Src has occurred, as is found in breast cancer, the proper negative regulation of this phosphorylation may be lost, resulting in the increased EGF-dependent signaling and tumorigenicity. We speculate that c-Src and EGFR act synergistically (via phosphorylation of the receptor by c-Src) to induce enhanced signaling in cells which overexpress both these kinases.