Phosphotyrosine Signaling Networks in Epidermal Growth Factor Receptor Overexpressing Squamous Carcinoma Cells

EXPERIMENTAL PROCEDURES

Preparation of Immobilized Antibody Affinity Resins—

Immunoaffinity resins (anti-pY and anti-EGF receptor) were prepared by covalent coupling to resin followed by low pH washing to remove noncovalently bound antibody, a potential source of significant immunoglobulin heavy and light chain contamination in later steps. Freshly prepared immunoaffinity resins were used for each biological experiment to maximize binding and reduce carryover. Anti-pY antibodies PY20 (Exalpha Biologicals, Inc., Watertown, MA) and PY100 (Cell Signaling Technology, Beverly, MA) were mixed in an 8:1 ratio and bound to Protein G-resin (Pierce, Rockford, IL) for 30 min at room temperature. In separate experiments, anti-EGF receptor antibody, recognizing an extracellular epitope (EGFR.1, Ref. 18; BD Bioscience, San Diego, CA), was similarly immobilized. Antibody resins were washed extensively, and 0.02 m dimethylpimelidate dihydrochloride (DMP, in 0.2 m triethanolamine, pH 8.2; Pierce) was added and mixed for 1 h on a rotating platform. Crosslinking was terminated by washing with three resin volumes of 0.2 m triethanolamine, pH 8.2, once with 0.2 m sodium citrate, pH 2.8 (to remove noncovalently bound IgG), and finally with three volumes of 10 mm TrisHCl, pH 8.2. Antibody crosslinked resins were stored at 4 °C until use. In later experiments disuccinimidyl suberate (DSS) crosslinker was substituted for DMP (Pierce) with similar results.

Preparation of HN5 Cell Lysate, Anti-pY, and Anti-EGF Receptor Affinity Chromatography and Protein Immunodetection—

Approximately 2 × 10⁸ HN5 cells (14) were grown in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum. HN5 cell extracts were prepared by mild detergent lysis (1% Triton X-100) containing protease and phosphatase inhibitors (see below) to enhance the preservation of protein interactions, lost when deoxycholate-containing lysis buffers (e.g. RIPA) were used. The selective EGF receptor kinase inhibitor erlotinib (1 μm OSI-774) was added to HN5 cells for 60 or 120 min prior to lysis. EGF-treated cells were incubated with ligand (10 ng/ml) for 10 min prior to lysis unless otherwise stated. Cells were washed once with PBS prior to lysis in 50 mm HEPES, pH 7.5, 150 mm NaCl, 1.5 mm MgCl₂, 1 mm EGTA, 10% glycerol, 1% Triton X-100, 1 mm 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 0.8 μm aprotinin, 20 μm leupeptin, 40 μm beestatin, 15 μm pepstatin A, 14 μm E-64 [1:100 dilution of protease inhibitor mixture P8340; Sigma, St. Louis, MO], sodium orthovanadate, sodium molybdate, sodium tartrate, and imidazole [1:100 dilution of phosphatase inhibitor mixture P5726; Sigma] for ∼3 min. Insoluable material was removed by centrifugation (13,000 × g, 10 min, 4 °C). Protein concentration was determined by microBCA assay (Pierce). Lysates were precleared by incubation with blank Protein-G resin for 30 min at 4 °C prior to immunoprecipitation to reduce nonspecific binding. Antibody resins were equilibrated with lysis buffer and incubated with HN5 cell lysates for 2–4 h at 4 °C with rotation. Antibody-antigen complexes were washed with >200 volumes of 10 mm TrisHCl, pH 8.0, 150 mm NaCl at 4 °C, proteins eluted with 0.1% TFA, 5% methanol, and dried in vacuo. The initial anti-pY affinity step yielded ∼50–100 μg of protein from 2 × 10⁸ cells, representing an approximate 1,000-fold enrichment, greatly reducing sample complexity for subsequent LC-MS/MS protein identification. Visual inspection of SDS-PAGE-fractioned anti-pY affinity-isolated proteins (Fig. 2A) or nonspecific proteins trapped on resin control (Fig. 2, B and C) or goat anti-rabbit antibody control (data not shown) qualitatively indicated minimal nonspecific binding and generally low IgG release from the resin. Equivalent amounts of protein extract, as determined by BCA assay, were subject to SDS-PAGE. Protein immunodetection was performed by electrophoretic transfer of SDS-PAGE-separated proteins to nitrocellulose, incubation with antibody, and chemiluminescent second step detection (ECL; Amersham, Piscataway, NJ).

Antibodies included: p130 Cas (#sc-860; Santa Cruz Biotechnology, Santa Cruz, CA), phospho-Shc (#2434; Cell Signaling, Beverly, MA), phospho-Paxillin (#2541; Cell Signaling), phospho-Akt (#9271; Cell Signaling), phospho-HER2/ErbB2 (#2245; Cell Signaling), phospho-p44/42 MAP kinase (#9101; Cell Signaling), phospho-EGFR (Tyr⁸⁴⁵) (#2231; Cell Signaling), phospho-EGFR (Tyr⁹⁹²) (#2235; Cell Signaling), phospho-EGFR (Tyr¹⁰⁴⁵) (#2237; Cell Signaling), EGFR (#2232; Cell Signaling), Pyk2 (#06–559-MN; Upstate Biotechnology, Lake Placid, NY), phospho-Caveolin-1 (#3251; Cell Signaling), phospho-p70 S6 kinase (#9205; Cell Signaling), phospho-GSK-3α/β (#9331; Cell Signaling), phospho-EGFR (Tyr¹⁰⁶⁸) (#2236; Cell Signaling), phospho-Src family (Tyr⁴¹⁶) (#2101; Cell Signaling), phospho-Src (Tyr⁵²⁷) (#2105; Cell Signaling), FAK (#06–543-MN; Upstate Biotechnology), Karyopherin B1 (#sc-11367; Santa Cruz Biotechnology), 14-3-3ζ (#sc-1019; Santa Cruz Biotechnology), 14-3-3ε (#610542; BD Transduction Laboratories, San Jose, CA), ACK (#sc-323; Santa Cruz Biotechnology), α-catenin (#sc-9988; Santa Cruz Biotechnology), Plakophilin 2 (#610788; BD Transduction Laboratories), GRB2 (#3972; Cell Signaling), δ-catenin (#07-259; Upstate Biotechnology), EF-1α (#05-235; Upstate Biotechnology), Tyk 2 (#sc-169; Santa Cruz Biotechnology), γ-catenin (#sc-8415; Santa Cruz Biotechnology), EphA2 receptor (#34-7400; Zymed Laboratories, South San Francisco, CA), clathrin heavy chain (#CP45; Oncogene Research Products, Boston, MA), c-Yes (#06-514; Upstate Biotechnology), and IGF-1R-β chain (#sc-713; Santa Cruz Biotechnology).

Peptide Identification by LC-MS/MS Fragment Ion Spectra Database Searching—

Proteins isolated by anti-pY affinity chromatography were denatured, reduced, carboxamidomethylated, and proteolytically cleaved with trypsin. Peptides were introduced into the Q-TOF mass spectrometer either using reverse phase (C18) HPLC or, to further reduce sample complexity and ion suppression, using coupled strong cation exchange-reverse phase (SCX-C18) HPLC. The use of multiple overlapping methods (Fig. 1) greatly improved the breadth of protein identification and peptide coverage, with greatest coverage with cleavable-ICAT and SCX-C18 strategies (data not shown). Two-dimensional SCX-C18 chromatography (19) was performed using a 1 × 5-mm cation exchange column packed with polysulfoethyl A resin (SCX; PolyLC, Columbia, MD) and 0.32 × 150-mm column packed with Pepmap C18 resin (LC Packings, San Francisco, CA) loaded and developed at 30 μl/min. Peptides were detected by UV absorbance at 214 nm using a 250-nl internal volume flow cell with a 2-mm path length. Peptides were eluted from the C18 resin then coupled directly to the mass spectrometer, at a flow rate of ∼2 μl/min. One-dimensional C18 chromatography was performed using a 0.1 × 150-mm column packed with C18 resins (MagicC18, Michrom Bioresources, Auburn, CA or Vydac MS218, Nest Group, Southborough, MA) and developed using a 2–70% ACN, 0.1% formic acid gradient with a flow rate of ∼500 nl/min. The electrospray source was fitted with an uncoated tapered fused silica tip (10–15-μm inner diameter; New Objective, Cambridge, MA) to which a voltage of 3.0 kV was applied with nebulizing nitrogen gas. Information-dependent MS and MS/MS acquisitions were made on an orthogonal Q-TOF (qQ-TOF) instrument (Sciex, Toronto, Canada) using a 1-s survey scan (m/z 400–1,200) followed by three consecutive 3-s product ion scans of 2⁺, 3⁺, and 4⁺ parent ions with a 4-min exclusion period. The product ion mass range was typically limited to 60–1,000 Da scanned in two cycles or 60–1,600 Da scanned in three cycles, and where collision energy was dynamically ramped according to mass and charge state, were used to maximize the duty cycle of the instrument. In later experiments, ions were stored in the second quadrapole and released in synchrony with the pulsing of ions in TOF detector, resulting in an ∼ 5–8-fold increase in sensitivity. Data were acquired using Sciex Analyst QS software. Proteins were identified from survey and product ion spectral data, with an MS and MS/MS mass tolerance of 0.15 Da, using both Swiss-Prot and Genbank NR databases and ProID (Version 1.0 EP2; Applied Biosystems, Foster City, CA), Mascot (Matrix Science, London, United Kingdom), and SONAR (Proteometrics, New York, NY) search programs. Mascot and SONAR searches accessed merged dta format files. Sciex wiff files were converted to dta format (“Export IDA Spectra” script) and merged using the Merge function from Matrix Science. Protein sequences for porcine trypsin, mouse immunoglobulin constant regions, and variable regions from the anti-pY antibody PY20 were appended to the human Swiss-Prot database. One missed tryptic cleavage was allowed and post-translational modifications considered included only cysteine derivitization and tyrosine phosphorylation. ProID confidence scores of >90% were considered, after which spectra were manually inspected as no criteria were found for any of the search programs that would allow correct unattended protein assignments without an unacceptably high false-negative rate. Mascot scores of >20 and SONAR expectation values of <1 were considered. Only proteins assigned with two or more peptides were included in Table I. Peptide redundancy was prevented by manually sorting the peptide lists in Excel.

The cleavable-ICAT labeling method was used to differentially measure protein abundance between two biological states, for example after EGF receptor blockade with erlotinib or after stimulation of receptor activity with EGF. After anti-pY affinity selection (described above), samples were dried in vacuo, denatured, and cysteine residues derivitized with either light isotope ICAT label (control group) or heavy isotope ICAT label (treatment group) according to the manufacturer’s protocol (Appled Biosystems, Foster City, CA) and subjected to information-dependent LC-MSMS as described above. The ratio between heavy and light labeled peptide provides a relative measure to the differential protein abundance within the anti-pY affinity fraction. ICAT-labeled proteins were identified using ProICAT (Version 1.0 SP3; Applied Biosystems), Mascot and SONAR, while the relative quantification of ICAT ion pairs was determined using ProICAT. ProICAT confidence scores of >90% were considered, after which spectra were inspected manually to confirm peptide sequence assignments. Multiple bioinformatics tools were used to visualize and analyze the connections between proteins and complexes identified by anti-pY affinity selection and LC-MS/MS spectra searching. Tools included Network Explorer (www.ingenuity.com) and ScanSite (scansite.mit.edu).

Protein Fractionation and Comparative MALDI-MS Peptide Mapping of Trypsin and gluC Digests—

Anti-pY affinity fractions were reduced (100 mm DTT) and carbamidomethylated (50 mm iodoacetamide, 1 h, room temperature, in the dark). Samples were fractionated by capillary C4 reverse-phase LC (0.32 mm × 150 mm) using a 5–90% gradient of increasing ACN, 0.1% TFA over 90 min with a flow rate of 3 μl/min. Reverse-phase protein fractions (∼5–20%, 20–30%, 30–40%, 40–50%, 50–60%, and 60–90% ACN) were reduced in volume and digested in 50 mm NH₄CO₃, 5% ACN with 50 μg/ml trypsin or 50 μg/ml gluC for 18 h at 37° C. Samples were desalted using micro reverse-phase C18 tips (Millipore Corp., Billerica, MA) and eluted in 50% ACN-water. Peptide masses were determined by MALDI (DE-Pro; Applied BioSystems, Framingham, MA) in reflector mode (2-m flight length) with a positive ion accelerating voltage of 20 kV, a grid voltage of 12.8 kV, guide wire voltage of 1,400 V, using 120 ns delayed extraction. Recrystallized α-cyano-4-hydroxycinnamic acid (4-HCCA) and dihydrobenzoic acid (DHB) were used as matrices, generally using dried droplet methods. Greater than 200 scans were averaged per spectra. Trypsin and gluC autodigestion products were used as internal mass standards. Protein prediction based on peptide mass information was performed by interrogation of the Swiss-Prot database using the Knexus search program (Proteometrics), searched at a resolution of 30 ppm for proteins from 10 to 300 kDa. Post-translational modifications considered included only cysteine modifications and phosphorylation of tyrosine residues.

RESULTS

Immunoaffinity Enrichment of Tyrosine-phosphorylated and EGF Receptor-associated Proteins and Complexes—

Phosphotyrosine-containing proteins and complexes were isolated and identified, using the HN5 HNSCC line as a model system. Constitutive EGF receptor signaling has been associated with altered signaling resulting in cell transformation and HN5 cells, expressing ∼5 × 10⁶ receptors per cell (20), show high-constitutive EGF receptor kinase activity due to autocrine TGF-α expression. From multiple biological experiments 774 unique peptides were identified and manually confirmed, corresponding to 236 proteins, not including trypsin, immunoglobulin, fetuin, or bovine serum albumin (∼27 peptides in total). The mean ProID/ProICAT peptide confidence and score were 99 and 29%, respectively. The 125 proteins identified with two or more peptides are listed in Table I and the peptide coverage of EGF receptor is listed in Table II. Protein assignments, confidence levels and scores for all peptides are shown in Supplemental Table I. Average errors in peptide mass measurements within information-dependent MS and MS/MS experiments using an qQ-TOF instrument were generally <0.05 Da and resolution values >9,000 allowed accurate charge state calculation. Nevertheless all assigned fragment ion peptide spectra were verified by visual inspection. Proteins were grouped using simplified Gene Ontology (GO; www.geneontology.org) process terms (Table I) where focal adhesion, microtubule, keratin filament, and actin filament proteins were bundled.

In HN5 cells the EGF receptor was the predominant phosphoprotein detected by anti-pY immunoblot (Fig. 2A), which was overexposed to reveal phosphorylated substrates whose phsophorylation could be prevented by inhibition of EGF receptor kinase activity. The EGF receptor showed the best peptide coverage by LC-MS/MS (defined by 80 peptides and 7 phosphopeptides; Table II) and by MALDI analysis (defined by 36 peptides; Table II). In this model system, EGF receptor represented the major scaffolding protein by which associated proteins and phosphoproteins were enriched. The three most-abundant classes of proteins observed (Table I and Fig. 3) were those associated with cell adhesion and cytoskeletal organization (43 proteins), mitogenesis and cell survival (23 proteins), and protein stability and degradation (17 proteins). In addition, two smaller-scale anti-EGFR affinity chromatography experiments also were performed where the monoclonal antibody (EGFR.1), recognizing an extracellular domain EGF receptor epitope, was used for protein selection. These experiments yielded 69 proteins in common with those identified by anti-pY affinity selection, and these are shaded gray in Table I.

EGF Receptor Phosphorylation—

Both known and unreported phosphopeptides from both unstimulated and EGF hyperstimulated cells could be readily identified from complex protein mixtures in both LC-MS/MS and MALDI workflows (Table III). These phosphopeptides were not detected when EGF receptor kinase activity was inhibited with erlotinib. Tyrosine phosphorylation of the four major autophosphorylation sites at positions Y1173 (P1), Y1148 (P2) (21), Y1114 and Y1086 (P4; Ref. 22) was observed within these complex peptide mixtures, with high ion counts and fragment ion coverage (Fig. 4). Phosphorylation of Y974 was observed (Table III) in both LC-MS/MS (2⁺ and 3⁺ ions) and MALDI studies (1⁺ ions), yielding a predicted SH2 interaction motif (ScanSite; scansite.mit.edu). Fragment ion spectra also indicated that S967 also could be phosphorylated within the same peptide, though co-phosphorylation of both sites was not observed. In vitro phosphorylation of anti-EGFR immune complexes with active Src kinase identified two additional EGFR phosphopeptides Y740 EILDEA(pY)VMASVDNPHVCR and Y703 VLGSGAFGTV(pY)K, though these could not detected directly from cell extracts. Immunoblot studies showed inhibition of EGF receptor autophosphorylation by erlotinib markedly decreased phosphophorylation of EGF receptor at sites Y845, Y992, Y1045, and Y1068 within the C-terminal SH2/PTB interaction domain (Fig. 5). ICAT labeling showed tyrosine phosphorylation at Y1086 and Y1148 in both the EGF treatment and no treatment control HN5 cells, but not after inhibition of EGF receptor kinase activity. The inhibition of EGF receptor kinase activity increased total EGF receptor levels (Fig. 5, left), most likely by attenuating autophosphorylation-dependent internalization.

EGF receptor can be tyrosine-phosphorylated at combinations of at least 12 sites in cells, and this high stoichiometry contributes to complexity in interactions, both with phosphotyrosine binding signaling partners and with the anti-pY affinity resin used for protein isolation. For example when erlotinib treatment (1 μm, 60 min) was followed by EGF stimulation (10 ng/ml, 10 min) prior to cell lysis and anti-pY affinity selection, marked down-regulation of EGF receptor in the anti-pY fraction was observed both by ICAT labeling (76 ± 4%; Table IV) and by immunoblot (Fig. 2A). However in the absence of exogenous EGF, little difference between control and erlotinib was observed by ICAT labeling (102 ± 4%; Table IV) in two biological experiments and multiple LC-MS/MS experiments. In the absence of EGF stimulation, the EGF receptor appeared to dephosphorylate relatively slowly and incompletely following erlotinib treatment, where sufficient tyrosine phosphate was present on EGF receptor to bind the pY affinity resin with similar efficiency as the control sample. This contrasted with rapid reduction in EGF receptor-interacting phosphoproteins such as Cbl-B, Erk1/2, PLCγ, and Vav-2 with erlotinib treatment in the absence of exogenous EGF (Table IV). This also contrasted with our finding with the constitutively active juxtamembrane mutant Kit receptor tyrosine kinase, where >80% dephosphorylation of the receptor rapidly occurs within 60 min of pharmacological Kit kinase inhibition, directly paralleling anti-pY capture of Kit (data not shown). These data suggest that whole-protein phosphotyrosine capture methods are also sensitive to the stoichiometry of protein phosphorylation and to the rates of dephosphorylation of the individual phosphorylation sites.

Activation and Inhibition of EGF Receptor Kinase Activity Modulates Proximal Signaling Components—

The relative abundance of phosphotyrosine-containing proteins and complexes from the HNSCC cell line HN5 were measured by immunoblot and by ICAT labeling from multiple biological experiments. Two types of experiments were performed. First, experiments were performed where EGF receptor kinase activity was inhibited by the selective EGFR kinase inhibitor erlotinib (1 μm for 60 or 120 min) with or without exogenous ligand (EGF 10 ng/ml for 10 min) prior to lysis and anti-pY affinity selection. In a second paradigm, EGF receptor was hyperstimulated with exogenous EGF ligand (10 ng/ml for 1, 3, or 15 min) prior to phosphoprotein isolation. In both cases, the cleavable ICAT cysteine labeling method was used to allow relative quantitation of proteins isolated under the different biological conditions. Treatment and control protein samples were separately labeled with heavy isotopic or light isotopic mass labels, respectively, degraded with trypsin and subjected to LC-MS/MS. ICAT labeling and subsequent recovery of only cysteine-derivitized peptides further reduced sample complexity. From erlotinib-treated or EGF-hyperstimulated cells, 33 proteins were measured by ICAT labeling and 26 proteins were measured by immunoblot. These data are summarized in Table IV. The majority of proteins were unaltered by more than 30% in representation after erlotinib (69% of proteins) or EGF (53% of proteins) exposure in vitro.

EGF receptor autophosphorylation establishes binding sites for proteins containing SH2 and PTB domains (7). Multiple SH2 domain-containing proteins were modulated by erlotinib and/or EGF exposure (Table IV), and these were modeled using the pathway connectivity Network Explorer software, from data obtained by anti-EGFR selection and from data obtained from the literature (Fig. 6). Competition between the anti-pY antibody and SH2/PTB phosphotyrosine binding proteins, which might restrict the detection SH2/PTB domain proteins, was not apparent. Vav2 and PLCγ were both acutely down-regulated by erlotinib, where only the D₀ control peak could be detected and sequenced. The converse was true for Vav2, where upon exposure to EGF only the D₉ treatment group labels were observed. The isolation of caveolin-1 in both anti-pY and anti-EGF receptor affinity fractions supports EGF receptor co-localization within caveolae. The activation of EGF receptor substrates PLCγ and PI-3 kinase substrates is thought to occur within lipid raft membrane compartments including caveolae (23). Consistent with this interpretation, caveolin-1 recovery within the anti-pY fraction was reduced by EGF receptor blockade (Fig. 7B). It has been proposed that PLCγ activation results in calcium mobilization within caveolae and subsequent activation of calcium dependent enzymes such as the S100 family members selected by anti pY affinity.

The MAP kinases Erk-1 and -2 are connected to EGF receptor signaling through the Ras-Raf-Mek pathway. As expected, kinase blockade reduced Erk representation by at least 68%, where the treatment D₉ peak was undetectable in two experiments (Table IV). The erlotinib reduction in Erk phosphorylation was qualitatively consistent with those obtained by immunoblot (Fig. 8). Hyperstimulation of EGF receptor activity by exogenous EGF increased Erk 200% somewhat late (15 min) after stimulation. The E3 ubiquitin ligase Cbl, which interacts with EGF receptor through the adapter Grb2, functions as a key negative regulator by promoting ubiquinylation and proteosomal destruction of the receptor. Both Cbl-E3 ligase and Cbl-B proteins were found in the anti-pY fraction, and Cbl-B was additionally observed in the anti-EGF receptor fraction (Table I). The abundance of Cbl-B in the phosphotyrosine fraction was reduced 73% by EGF receptor kinase inhibition and was increased by EGF treatment 160 and 240% after 1 and 15 min exposure, respectively. Additional components of internalization and proteosomal degradation complexes were identified, including the 26S proteosomal regulatory subunit-2 (by both LC-MS/MS and immunoblot), the F-box only protein-2, S100A6, calcyclin binding protein and ubiquitin (by LC-MS/MS), and dynamin, cullins 1 and 2, clathrin heavy chain, and the ubiquitin C-terminal hydrolase (by immunoblot; data not shown).

Alternative approaches to the dissection of EGF receptor signaling have been reported (24–27). Stable isotope-labeled proteins from the EGF-stimulated cervical carcinoma cell line HeLa (∼2 × 10⁵ receptors/cell) were selected by Grb2-SH2 domain affinity, fractionated by SDS-PAGE, and peptides subjected to LC-MS/MS. The Grb2-SH2 domain shows preference for pYKNI/L, is permissive for pYXNX, and interacts with EGF receptor at multiple positions. Due to the involvement of Grb2 in EGF receptor signaling, we examined the cross-identification of proteins between the two studies. Of the 228 proteins identified by Grb2-SH2 interaction in HeLa cells (24), ∼49 (21%) were common in the full phosphotyrosine affinity fractions characterized here (Supplemental Table I). Of those proteins modulated by ligand stimulation of EGF receptor, Shc, Grb2, Vav-2, Cbl-B, polyubiquitin, actin, keratin-17, and plectin-1 were common between the two studies. A previous study using phosphotyrosine precursor ion scanning of HeLa cell anti-pY selected proteins, separated by SDS-PAGE, identified 10 proteins, four of which were common to the study here (EGF receptor, Cbl, Hsp70, and Shc; Ref. 25). Similarly, anti-pY and two-dimensional gel electrophoretic approaches to mapping of A431 cell proteins yielded 16 proteins overlapping with those reported here (comprising EGFR, PLCγ, α₆-integrin, γ-catenin, PI-3 kinase p85β, ezrin, Grp78, cortactin, Hsp71, Grp75, α-tubulin, Shc, actin, and Grb2) out of 19 proteins in total (26). These differences suggest diversity between the HeLa, A431, and HN5 cell systems and the different methodologies employed.

Tyrosine Phosphorylation Contributed by Non-EGF Receptor Protein Kinases—

Twelve tyrosine protein kinases and three serine-threonine protein kinases, from multiple functional categories, were among the proteins isolated by anti-pY) affinity chromatography. Despite the dominant role of EGF receptor in the model cell system used, FAK (19 peptides), ErbB2 (7 peptides), c-Yes (7 peptides), c-Src (6 peptides), IGF-1 receptor (2 peptides), Pyk2 (3 peptides), EphA2 (7 peptides), EphB3 (3 peptides), EphB4 (2 peptides), and ACK1 (4 peptides) likely also influence the proteins selected by anti-pY affinity chromatography. ErbB2, Yes, FAK, EphA2, EphB4, and ACK1 also were also observed by anti-EGFR affinity selection.

As a consequence, we considered the relative contributions of these tyrosine kinases to the proteins and complexes isolated and considered the extent of cross-talk between kinases. For example activation of EGF receptor, by ligand or by mutation, has been shown to result in the activation of Src, through indirect phosphorylation of the activating Src Y419 site within the kinase activation loop. In turn, activated Src is known to phosphorylate EGF receptor on Y845 and Y1101 (28), thereby recruiting signaling proteins. The activation of Src and Yes kinases was indicated by several pieces of evidence. First, EGF receptor was phosphorylated on Y845, a known Src phosphorylation site (Fig. 5). Second, Src phosphorylation at Y419, required for full kinase activity, was observed by immunoblot (Fig. 7B) and was strongly suggested by LC-MS/MS (data not shown). Third, p130CAS was phosphorylated on multiple Src family sites with the central tyrosine-rich domain (Table III). LPA is a mitogen known to use Src and EGF receptor to transduce signals initiated by LPA binding to its G-protein coupled receptors (28–31). LPA addition to HN5 cells increased phospho-Erk and phospho-Akt (Fig. 8), which could be blocked by the selective EGF receptor kinase inhibitor erlotinib. Erlotinib treatment reduced phosphorylation of the EGF receptor at Y845, a site phosphorylated by Src kinases, consistent with a cross-activation of EGF receptor and Src family members (32, 33). However, no change in phospho-Src family Y419 (active state; Fig. 7B), phospho-Src Y530 (repressed state), or anti-pY Yes could be detected by immunoblot even after 120 min of EGF receptor kinase inhibition (data not shown). These data suggest a more rapid dephosphorylation of EGF receptor Y845 relative to Src Y419, or distinct compartmentalization of the EGF receptor and Src family kinases (8).

IGF-1 receptor was identified by anti-pY selection from HN5 cells and exogenous IGF-1 increased phospho-Akt and phospho-GSK3, but antagonized phosphorylation of Erk (Fig. 8). However, the phosphorylation of Shc was still effectively blocked by erlotinib (OSI-774), indicating its phosphorylation was largely EGF receptor-dependent and was not perturbed by LPA or IGF-1. Interestingly, the IGF-1-dependent phosphorylations of GSK-3β and Akt could not be overcome through EGF receptor inhibition, indicating IGF-1 activation of the PI-3 kinase pathway was insensitive to EGF receptor blockade. LPA stimulated phospho-paxillin levels while erlotinib modestly decreased paxillin phosphorylation (Figs. 7B and 8). No alteration in β-arrestin was observed under these conditions (Fig. 8).

Modulation of Cell Adhesion Complexes by Activation and Inhibition of EGF Receptor Kinase Activity—

Cell adhesion proteins including Crk-associated substrate (p130CAS), focal adhesion kinases (FAK and Pyk), and paxillin were readily observed in anti-pY LC-MS/MS and immunoblot experiments. These proteins were shown to form stable interactions with EGF receptor by anti-EGFR affinity selection. Both paxillin and p130CAS have been shown to directly interact with FAK (34), as part of a complex with actin required for focal adhesion assembly. Two phosphopeptides from paxillin were observed (Table III), comprising one known site at Y118 and one unreported site at Y88, N-terminal to the FAK-binding domains and LIM domains. The major paxillin phosphorylation site at Y31 is resident within a peptide mass of 6,031 Da, somewhat too large to be readily detected in the LC-MS/MS scheme used. p130CAS contains an N-terminal FAK-interacting SH3 domain, a centrally located substrate-interacting (Crk) repeat region with multiple tyrosine residues, followed by a C-terminal Ser region responsible for Src binding. Within the central region, tyrosine phosphorylation was unequivocally observed at sites Y249, Y306, and Y387 in the motif YDVP (Table III) and was strongly suggested for Y410 within the motif YAVP (data not shown). The tyrosine phosphorylation of the YDVP motif has been shown to be phosphorylated in vitro by Src (35), while direct phosphorylation of this motif by EGF receptor has not been reported.

The assembly of focal adhesion complexes containing FAK/Pyk2, paxillin, p130CAS, and ARF-GIT1 is thought to be regulated through phosphorylation and dephosphosphorylation in a closely balanced manner (13). The anti-pY capture of these focal adhesion proteins was markedly decreased following hyperactivation of the EGF receptor by exogenous EGF. For example in HN5 cells treated with EGF, p130CAS was decreased 21, 49, and 29% after 1, 3, and 15 min, respectively. ARF GTPase-activating protein GIT1 was decreased by EGF treatment 36 and 43% after 3 and 15 min and FAK was decreased 51% after a 1 min (Table IV). A decrease in focal adhesion complex phosphorylation following EGF addition is consistent with data obtained in A431 cells and indicate that despite EGF receptor-mediated phosphorylation of focal adhesion proteins, the phosphotyrosine content of the complex was decreased due to the rapid phosphatase activation and dephosphorylation of focal adhesion complexes (13). Protein complexes were modeled using the pathway analysis Network Explorer software and through literature data (Fig. 7A). Data from ICAT and immunoblot experiments were superimposed upon this framework to visualize points of regulation (Fig. 7B).

Interestingly, EGF receptor kinase inhibition for 120 min elicited a similar more modest pattern of decreased recovery of focal adhesion regulating proteins within the phosphotyrosine fraction (Table IV and Fig. 7). For example, erlotinib reduced anti-pY capture of FAK1 by 58%. Similarly, immunoblot experiments indicate p130CAS was reduced in the anti-pY fraction but not in the anti-EGF receptor fraction, consistent with a reduction in tyrosine phosphorylation state but not in EGF receptor interaction (Fig. 7B), which was underestimated by ICAT labeling (Table IV). Similarly, anti-pY capture of paxillin, p130CAS, and the FAK homolog Pyk2 were all reduced by EGF receptor inhibition (Fig. 7B), while no changes in the phosphorylation of ACK1, α-catenin, or Tyk2 (Figs. 6 and 7B) were observed. In HN5 cells, transient EGF receptor inhibition directly or indirectly reduced phosphorylation of cell adhesion components.

DISCUSSION

The sustained activation of the EGF receptor tyrosine kinase, through mutation or autocrine and paracrine stimulation, has been shown to promote phenotypic transformation in vitro and in vivo (36–38). The biological experiments performed here, in a model of HNSCC, have identified proteins tyrosine-phosphorylated predominantly by EGF receptor and Src family kinases, and proteins that form stable associations with tyrosine-phosphorylated proteins, comprising multiple functional classes (Fig. 3 and Table I). The anti-pY affinity capture method enabled a broad view of EGF receptor and cell adhesion signaling under conditions of EGF receptor autocrine activation, of hyperactivation with exogenous ligand, and of kinase inhibition. Antibody capture methods can suffer from an unacceptable level of nonspecific binding, confounding the identification of proteins specifically interacting with a given target. The use of both an EGF receptor kinase inhibitor and EGF stimulation allowed us to discern pharmacologically and physiologically regulated events, relatively insensitive to the effects of nonspecific binding. We also used several approaches to minimize nonspecific binding, including preabsorption to blank affinity resin, the use of covalently bound pH 3 washed antibody resin, and the use of 0.1% Triton in the initial wash steps. Multiple phosphorylation sites, both novel and known, could be mapped despite the complexity of the peptide mixtures. In proteins showing a sharp treatment-related decrease in anti-pY selection (e.g. Erk and Vav-2), the dynamic range of ICAT measurements was sometimes limited by baseline noise within MS spectra. Thus the combined use of the higher-throughput ICAT approach to identify biologically modulated proteins, followed by immunoblotting to verify and more accurately measure protein abundance, was useful. Multiple biological and LC-MS/MS experiments were performed for both protein identification and for ICAT quantitation. The average error in ICAT measurements between repeat experiments (Table IV) was 6%. While the reproducibility in quantitation of peptides between experiments was high, the reproducibility in peptide identification between experiments was considerably more variable. This is likely due to the relatively high degree of sample complexity and the potential for variable ion suppression between LC-MS/MS and biological experiments. Ion suppression effects would likely be improved with additional fractionation by SCX prior LC-MS/MS.

EGF Receptor Immediate Signaling Complexes and Receptor Cross-talk—

HN5 cells were shown to utilize multiple SH2 scaffolding proteins NSP1, NSP2, Grb2, Shc, PI3-kinase, PLCγ, STAT3, c-Src, c-Yes, Cbl E3-ligase, Cbl-B, Vav-2, and possibly Tyk2 to establish signaling complexes. NSP1, NSP2 (39), and Vav2 (40, 41) can directly interact with EGF receptors and function as guanine nucleotide exchange factors (GEFs) for small G-proteins of the Ras and Rho family, respectively. Phosphorylation of the mature EGF receptor from HN5 cells was observed on tyrosines Y845, Y992, and Y1068 (by immunoblot) and Y974, Y1086, Y1114, Y1148, and Y1173 (by LC-MS/MS), establishing SH2 and PTB interaction sites. In vitro addition of 1 mm ATP and active Src to EGF receptor immunocomplexes identified Y703, Y740 as additional tyrosine phosphorylation sites (data not shown), but it is not known whether such sites can be utilized in vivo. However, synthetic phosphopeptide binding studies do highlight Y703, Y740, Y992, and Y1173 as Shc interaction motifs and Y740, Y1068, Y1086, Y114, and Y1148 as Grb2 interactions sites (42).

Complex assembly mediated by SH2 and PTB domain interaction is important not only for the generation of mitogenic and survival signals, but also for the regulation of the rate and route of EGF receptor internalization and consequent signal termination. For example, the SH2 domain-containing E3 ligases, Cbl and Cbl-B, both present within the anti-pY and anti-EGFR fractions, can interact with EGF receptor either directly via SH2 interaction or via SH3 interaction through the SH2-containing adapter Grb2 and are required for efficient proteosomal degradation of the receptor (43–45). Grb2 further interacts with dynamin to promote the transition of internalization complexes from coated pits to endosomal vesicles. Disruption of Cbl function, for example in v-Cbl, transforms cells in part by rerouting activated receptor tyrosine kinases, including EGF receptor, back to the cell surface and promoting their escape from proteosomal degradation. The actions of Cbl receptor degradation are quite specific for the individual erbB family members. EGF receptor homodimers are efficiently down-regulated and degraded in response to Cbl binding, while EGF receptor-ErbB2 heterodimers show a greater degree of recycling back to the cell membrane and thus exhibit enhanced cellular transforming activity (46). The endosomal sorting signal generated by Cbl requires EGF receptor kinase activity (45) and the phosphorylation sites within the C-terminal tail (47), which may explain the accumulation of total EGF receptor observed after kinase inhibition (Fig. 5).

Members of the Src family of nonreceptor tyrosine kinase have been shown to influence cell proliferation, survival, cell adhesion, and migration, and Src has been shown to cooperate with EGF receptor to enhance cellular transformation, independently of receptor ligand binding (28). Both Src and Yes directly interact with and can be activated by EGF receptor (39), and Src was recently shown to contribute to constitutive phosphorylation and activation of STAT3 in HNSCC cell lines, independently of JAK kinases (48). In HN5 cells, the phosphorylation of EGF receptor on pY845 (49) and the phosphorylation of Src on pY419, a site required for Src kinase activation, indicate Src family kinases were in an active conformation and can contribute to downstream signaling. In contrast to Src and IGF-1 receptor, where phosphorylation of the respective kinase loops is required for full kinase activation, phosphorylation of EGFR Y845 has little effect on kinase activity (50). Recent data suggest Y845 may function as a docking site for downstream effectors (51). Inhibition of EGF receptor kinase activity reduced Src family phosphorylation at Y845, consistent with findings in other cell types (39). At physiological ATP concentrations, this quinazoline class of EGF receptor kinase inhibitor has no detectable direct effect on Src kinase activity (52), indicating Src family activation was in part dependent on EGF receptor kinase activity. While EGF receptor blockade inhibited Src/Yes mediated EGFR Y845 phosphorylation (Fig. 5) and LPA phosphorylation of Erk, Akt and S6 (Fig. 8), the phosphotyrosine Y419 signal from the Src family showed little change for up to 120 min following EGF receptor kinase inhibition. These data suggest that while EGF receptor associated Src family activity may be down-regulated in response to receptor kinase inhibition, while the total pool of Src family activity remained unchanged, suggesting compartmentalization of different Src “pools.” Cross-talk between the IGF-1 receptor and EGF receptor also was observed (Fig. 8), where EGFR kinase blockade attenuated the IGF-1 stimulation of Erk and ribosomal S6 phosphorylation (Fig. 8). This is in agreement with data from HB4A cells, where EGF receptor blockade down-regulated IGF-1-stimulated Erk phosphorylation (53). Interestingly, IGF-1 stimulation reduced basal Erk phosphorylation in EGF receptor-overexpressing HN5 cells, suggesting competition between the two receptors for access to components of the Ras-Raf-MEK pathway.

Cell Adhesion Complexes—

The control of actin filament formation and cell-cell contacts, associated with cell adhesion, cell migration, and mitosis, are critical to the maintenance of normal cell function. Multiple members of focal adhesion and cell-cell junction complexes were identified by anti-pY and anti-EGFr selection and LC-MS/MS approaches. Of these, FAK, Pyk2, p130CAS, ARF-GIT1, and paxillin were modulated by activation and inhibition of EGF receptor kinase activity. Ligand-activated EGF receptor forms complexes with the C-terminal domain of FAK and Y397 in the N-terminal domain of FAK is phosphorylated upon EGF stimulation. FAK is required for EGF-stimulated cell migration, though its kinase activity is not obligatory (12). Paradoxically, hyperactivation of the EGF receptor by EGF addition lead to a time-dependent decrease in the recovery of p130CAS, Pyk, paxillin, and FAK in the phosphotyrosine affinity fraction (Table IV). However, this may be explained by the reciprocal relationship between EGF receptor activity and focal adhesion formation involving FAK, p130CAS, and paxillin (13). Studies by Lu and coworkers indicate that growth factor receptor activation recruits a tyrosine phosphatase, likely SHP, to dephosphorylate and disrupt focal adhesion complexes, thereby allowing cell migration. When the EGF receptor kinase was transiently inhibited by erlotinib, a similar pattern of focal adhesion phosphoprotein decrease was observed, for example for FAK as measured by ICAT labeling and for Pyk and paxillin by immunoblot (Fig. 7B). This also was observed for p130CAS, where isolation within the phosphotyrosine affinity fraction was decreased by EGF receptor blockade (Fig. 7B), while interaction with EGF receptor was unchanged. The data indicate transient EGF receptor inhibition also leads to decreased phosphorylation of focal adhesion components, suggesting a close balance between the phosphorylation of focal adhesion components and tyrosine phosphatase activation.

The catenins and plakophilins are members of the armadillo family of proteins and serve in the assembly of cadherin and desmosomal complexes important in cell-cell interaction (54). Tyrosine phosphorylation of β-catenin (Y654) by EGF receptor has been shown to disrupt the complex linking E-cadherin to α-catenin and actin filaments (55–57), and the decreased expression of α-catenin and E-cadherin has been proposed to contribute to the aggressiveness of basaloid squamous carcinoma (58). Following inhibition of EGF receptor kinase activity, the only changes occurring in these complexes were a modest decrease in phosphorylation and/or association of plakophilin. However, the presence of CD98, integrin α₆β₄, and bullous pemphigoid antigen (BPAG) within the phosphotyrosine fraction was of interest. The cell-surface integrin α₆β₄ functions as a adhesive receptor for the basement membrane laminins (59) and forms complex(s) with BPAG, plectin, and keratin filaments, which are dissolved by EGF receptor activation to allow migration (60). CD98 has been reported to be constitutively associated with integrin and functions to cluster and activate integrins resulting in PI3-kinase and Akt activation and anchorage-independent growth (61). While activation of integrin-linked kinase (ILK) can lead to tyrosine phosphorylation of integrin β₄ and subsequent activation of Erk and PI3-kinase mitogenic and survival signaling pathways, both integrin α₆β₄ and EGF receptor have been shown to directly associate, also promoting integrin β₄ phosphorylation possibly through associated Src family kinases Fyn and Yes. Conversely, integrin receptor complexes can activate EGF receptor in a Src- and p130CAS-dependent manner, leading to EGF receptor phosphorylation at Y845, Y1068, Y1086, and Y1173 (62). While no modulation of integrin or CD98 by EGF or erlotinib was observed, it is formally possible for these complexes to use EGF receptor as a signaling scaffold in a receptor kinase-independent manner. These data suggest additional mechanisms by EGF receptor might serve as a signaling scaffold independent of its kinase activity.

Rapid methods for identifying EGF receptor-associated pathway constituents and semi-quantitatively assess the impact of EGF receptor kinase activity on protein phosphorylation and complex formation were established. Proteins and complexes with known and poorly described functions were unequivocally identified by anti-pY and anti-EGFR affinity selection and pathways modeled. Specific interactions between EGF receptor, adaptor proteins, downstream signaling components, and interacting kinases were observed that could be modulated by EGF receptor kinase activity. Protein interactions within these complexes can be further refined by specific antibody selection experiments or by tandem affinity selection by expression of epitope-tagged cDNAs (63, 64). While the activation of TCF-LEF, STAT3/5, and AP-1 are relatively well studied, the role for EGF receptor kinase as a nuclear factor directly contributing to transcription has only recently been documented (65, 66). The recovery of nuclear transport and transcription components, though suggestive of a direct role for the EGF receptor kinase in transcription, also may be explained by nuclear localization of phosphorylated STAT3 and its known association tranporter protein importin-β1 (67). Functional interactions between EGF receptor, IGF-1 receptor, Src, and LPA signaling converged in downstream Mek-Erk and PI-3 kinase pathways and suggest EGF receptor can additionally serve as a ligand-independent signaling scaffold for distinct kinases. The recent development of methods for multiplex isobaric peptide labeling provide increased sensitivity and dynamic range, enabling time and dose studies in sensitive and resistant tumor lines and xenografts (68). The application of these methods to the further examination of EGF receptor and cell adhesion signaling will allow time-dependent quantitation of phosphotyrosine-containing proteins complexes following kinase activation and inhibition between tumor cells responsive and insensitive to EGF receptor blockade.