Regulation of epidermal growth factor receptor signaling by phosphorylation of the ras exchange factor hSOS1.

In response to stimulation with epidermal growth factor (EGF), the guanine nucleotide exchange factor human SOS1 (hSOS1) promotes the activation of Ras by forming a complex with Grb2 and the human EGF receptor (hEGFR). hSOS1 was phosphorylated in cells stimulated with EGF or phorbol 12-myristate 13-acetate or following co-transfection with activated Ras or Raf. Co-transfection with dominant negative Ras resulted in a decrease of EGF-induced hSOS1 phosphorylation. The mitogen-activated protein kinase (MAPK) phosphorylated hSOS1 in vitro within the carboxyl-terminal proline-rich domain. The same region of hSOS1 was phosphorylated in vivo, in cells stimulated with EGF. Tryptic phosphopeptide mapping showed that MAPK phosphorylated hSOS1 in vitro on sites which were also phosphorylated in vivo. Phosphorylation by MAPK did not affect hSOS1 binding to Grb2 in vitro. However, reconstitution of the hSOS1-Grb2-hEGFR complex showed that phosphorylation by MAPK markedly reduced the ability of hSOS1 to associate with the hEGFR through Grb2. Similarly, phosphorylated hSOS1 was unable to form a complex with Shc through Grb2. Thus phosphorylation of hSOS1, by affecting its interaction with the hEGFR or Shc, down-regulates signal transduction from the hEGFR to the Ras pathway.

In response to stimulation with epidermal growth factor (EGF), the guanine nucleotide exchange factor human SOS1 (hSOS1) promotes the activation of Ras by forming a complex with Grb2 and the human EGF receptor (hEGFR). hSOS1 was phosphorylated in cells stimulated with EGF or phorbol 12-myristate 13-acetate or following co-transfection with activated Ras or Raf. Co-transfection with dominant negative Ras resulted in a decrease of EGF-induced hSOS1 phosphorylation. The mitogen-activated protein kinase (MAPK) phosphorylated hSOS1 in vitro within the carboxyl-terminal proline-rich domain. The same region of hSOS1 was phosphorylated in vivo, in cells stimulated with EGF. Tryptic phosphopeptide mapping showed that MAPK phosphorylated hSOS1 in vitro on sites which were also phosphorylated in vivo. Phosphorylation by MAPK did not affect hSOS1 binding to Grb2 in vitro. However, reconstitution of the hSOS1-Grb2-hEGFR complex showed that phosphorylation by MAPK markedly reduced the ability of hSOS1 to associate with the hEGFR through Grb2. Similarly, phosphorylated hSOS1 was unable to form a complex with Shc through Grb2. Thus phosphorylation of hSOS1, by affecting its interaction with the hEGFR or Shc, down-regulates signal transduction from the hEGFR to the Ras pathway.
The guanine nucleotide exchange factor human SOS1 (hSOS1) 1 (1), the homologue of Drosophila Son of sevenless, promotes the activation of Ras following stimulation with epidermal growth factor (EGF) or with other growth factors acting through receptor tyrosine kinases (2). The molecular mechanism of Ras activation by EGF involves the interaction of hSOS1 with the activated human EGF receptor (hEGFR) through the adapter protein Grb2 (3). Grb2 links the hEGFR to hSOS1 by binding phosphotyrosine 1068 on the cytosolic tail of the hEGFR with its SH2 domain, and the carboxyl-terminal proline-rich region of hSOS1 with its SH3 domains (4,5). It has been suggested, that upon formation of the hSOS1-Grb2-hEGFR complex, hSOS1 is recruited to the plasma membrane where it activates Ras by promoting GDP release and GTP binding (6,7).
Stimulation with EGF also triggers the interaction of the hEGFR with the Shc proteins. The mammalian Shc gene encodes three proteins of 46, 52, and 62 kDa, and upon treatment with EGF Shc binds the activated hEGFR and undergoes tyrosine phosphorylation (8). Phosphorylation of Shc on tyrosine residue 317 creates a binding site for the SH2 domain of Grb2 and promotes the formation of the complex hSOS1-Grb2-Shc which participates in the activation of Ras (9), (10). Active, GTP-bound Ras binds and causes the translocation of the serine threonine kinase Raf to the plasma membrane where Raf is activated by an event as yet unidentified (11,12). Active Raf phosphorylates Mek which phosphorylates and activates the mitogen-activated protein kinases (MAPK) Erk1 and Erk2 (13). MAPK phosphorylates a number of cytosolic kinases and nuclear transcription factors which contribute to elicit the cellular responses following growth factor stimulation (14). EGF induced activation of the Ras signaling pathway is often short lived, Ras⅐GTP loading and MAPK activation reach a maximum within 2-5 min and return to a basal level in 1 h (15). Several mechanisms which cooperate to the down-regulation of Ras signaling have been described. Among these mechanisms phosphorylation by MAPK or by other serine/threonine kinases acting downstream of Ras results in desensitization of the hEGFR (16) and in down-regulation of Raf and Mek (17). In cells treated with EGF or insulin hSOS1 as well as its murine counterpart, mSOS1 (18), undergoes serine/threonine phosphorylation (19,20). These observations, and the fact that MAPK phosphorylates Drosophila SOS in vitro on sites which are also phosphorylated in vivo (21), have suggested that MAPK phosphorylates hSOS1 in response to growth factors activating receptor tyrosine kinases. It has been speculated that phosphorylation of hSOS1 by MAPK constitutes a negative feedback mechanism participating in the down-regulation of Ras signaling (22).
Here we show that active MAPK phosphorylates hSOS1 in vitro within the carboxyl-terminal proline-rich domain, on sites which are also phosphorylated in cells stimulated with EGF. Phosphorylation by MAPK does not affect hSOS1 interaction with Grb2 in vitro, nevertheless, phosphorylated hSOS1 demonstrates a markedly decreased ability to form a complex with the hEGFR or Shc through Grb2. Although other serine/threonine kinases cooperate with MAPK in hSOS1 phosphorylation in vivo, our data suggest that hSOS1 phosphorylation participates in the down-regulation of EGF induced activation of Ras.

MATERIALS AND METHODS
COS1 Cell Transfection-The cDNAs of interest (10 g each), cloned into the pEXV3 expression vector, were transfected into COS1 cells by electroporation (23). COS1 cells, grown to 75% confluence, were harvested and washed twice in HEBS buffer (20 mM HEPES pH 7.05, 137 mM NaCl, 5 mM KCl, 0.7 mM Na 2 HPO 4 , 6 mM glucose). 2.5 ϫ 10 6 cells were resuspended in 240 l of HEBS and placed into a 0.4-cm Bio-Rad electroporation cuvette containing 100 g of sonicated salmon testes DNA (10 l) (Sigma) and the plasmid DNA, to make a total volume of 260 l. Cells were electroporated at 250 V/125 microfarads giving a time constant of 5 to 6 ms. Following electroporation the cells were seeded * This work was supported by the Bayer Corp. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Immunoprecipitation and Immunoblotting-An identical protocol was used for COS1 and HER14 cells. Cells were serum starved for 16 h, treated with EGF and/or phorbol 12-myristate 13-acetate (PMA), and lysed in 1 ml/dish of ice-cold 20 mM Tris-HCl, pH 8, 135 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 1% Triton X-100, 10% glycerol, 1 mM Pefabloc, 20 g/ml each of leupeptin, aprotinin, soybean trypsin inhibitor, 50 mM NaF, 1 mM NaVO 4 . Lysates were cleared of the insoluble material, normalized for protein concentration, and immunoprecipitated at 4°C for 2 h using the required antibody. Protein A-Sepharose beads were then added for a further hour to recover the immunoprecipitates. Immunoprecipitates were washed 3 times with 1 ml of lysis buffer, resolved by SDS-PAGE (6% gels, Novex, San Diego, CA) and, when required, electrophoretically transferred to a poly(vinylidene difluoride) membrane (Millipore, Bedford, MA). Membranes were incubated with the required antibody and immunoreactive proteins were visualized using enhanced chemiluminescence following incubation with a secondary antibody conjugated to horseradish peroxidase. To strip the membranes of the prebound antibody, membranes were incubated in 200 mM glycine, pH 2.2, 0.1% SDS, 1% Tween 20, for 1 h at room temperature. For metabolic labeling, COS1 cells, transfected with the required plasmids, were serum starved for 16 h and incubated for a further 4 h in phosphate-free Eagle's medium (2 ml/60 mm plate) containing 0.5 mCi/ml of 32 P i . Cells were lysed and the proteins of interest were immunoprecipitated, resolved on SDS-PAGE, and visualized by autoradiography.
In Vitro Phosphorylation of hSOS1-Bacterially produced GST-Erk1 was incubated with active Raf and recombinant Mek for 30 min at 30°C in 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 50 M ATP (24). Under these conditions active Raf, which was purified from Sf9 cells co-infected with c-raf, v-ras, and v-src baculoviruses, phosphorylated Mek which in turn phosphorylated and activated GST-Erk1 (24). An aliquot of the Raf-Mek-GST-Erk1 kinase reaction was then mixed with the required amount of baculovirus produced, epitope (EYMPME, Glu-Glu)-tagged hSOS1 (25) and incubated at 30°C for a further 30 min in the same reaction buffer. When required [␥-32 P]ATP (6000 Ci/mmol, 10 mCi/ml) was added to the kinase reaction at the final concentration of 0.5 Ci/l. Typically 3 g of GST-Erk1, which was pre-activated by incubation with 0.5 g of Raf and 1 g of Mek, were used to phosphorylate 10 g of recombinant hSOS1 in a 100-l reaction.
Two-dimensional Tryptic Peptide Mapping-COS1 cells, transiently expressing epitope-tagged (Glu-Glu, EYMPME) hSOS1, were metabolically labeled with 32 P i and hSOS1 was immunoprecipitated using an anti-Glu-Glu mAb. At the same time 10 g of recombinant hSOS1 were phosphorylated in vitro using activated GST-Erk and [␥-32 P]ATP. hSOS1 was resolved by SDS-PAGE and electrophoretically transferred onto a poly(vinylidene difluoride) membrane. The hSOS1 band was visualized by autoradiography, cut out, and incubated for 12 h with 15 g of trypsin in 70 l of 50 mM ammonium bicarbonate, pH 8, at room temperature (26). Following the first 12 h of incubation a further 15 g of trypsin were added and the incubation was continued for another 6 h. The tryptic peptides were lyophilized twice in water and once in pH 1.9 electrophoresis buffer (formic acid/acetic acid/water, 1/3/36, v/v). Lyophilized peptides were resuspended in 12 l of electrophoresis buffer and 10,000 cpm, typically corresponding to 2-6 l, were loaded to 20 ϫ 20-cm glass-backed cellulose plates. Peptides were separated in two dimensions by electrophoresis at 1 kV for 1.5 h using pH 1.9 electrophoresis buffer followed by chromatography (2-butanol/pyridine/acetic acid/water, 6.4/5/1/4) and autoradiographed (27).
Reconstitution of the hSOS1-GST-Grb2 Complex-Recombinant hSOS1 was phosphorylated in vitro using activated GST-Erk and, when required, [␥-32 P]ATP. Bacterially produced GST-Grb2 (10 pmol) was mixed with 32 P-labeled phosphorylated hSOS1 (6.5 pmol) and the required amount of unmodified or phosphorylated hSOS1 in 200 l of lysis buffer. Tubes were rotated at 4°C for 15 min then the GST-Grb2-hSOS1 complexes were recovered using glutathione-agarose beads. Beads were washed 3 times, complexes were resolved on SDS-PAGE, and the radioactivity bound to GST-Grb2 was quantified using an AMBIS ␤ scanner. Under these conditions the amount of 32 P-labeled phosphorylated hSOS1 remaining bound to GST-Grb2 estimated the ability of unmodified hSOS1 and of phosphorylated hSOS1 to form a complex with GST-Grb2.
Reconstitution of the hSOS1-GST-Grb2-hEGFR Complex-hEGFR was immunoprecipitated using an anti-hEGFR mAb (EGFR1) (28) from HER14 cells (29), which were serum starved for 16 h and treated with EGF or PMA as required. Immunoprecipitates were immobilized onto Protein A-Sepharose and washed 3 times with lysis buffer. hEGFR-Protein A-Sepharose beads were incubated with 3 pmol of epitope (Glu-Glu)-tagged unmodified hSOS1 or in vitro phosphorylated hSOS1 and 3 pmol of GST-Grb2 in 1 ml of lysis buffer. Following 1 h at 4°C the beads were washed and the formation of the hEGFR-GST-Grb2-hSOS1 complex was detected by analyzing the immunoprecipitates on SDS-PAGE followed by immunoblotting with an anti-Glu-Glu mAb (30). To ascertain that equal amounts of hEGFR immunoprecipitates were used in each condition, blots were stripped of the anti-Glu-Glu antibody and incubated with an anti-hEGFR mAb (Transduction Laboratories, Lexington, KY). We ensured that an equal amount of unmodified and phosphorylated hSOS1 was added to each reconstitution assay by analyzing equivalent amounts of unmodified and phosphorylated hSOS1 on SDS-PAGE followed by Western immunoblotting.
Reconstitution of the hSOS1-GST-Grb2-Shc Complex-Shc was immunoprecipitated from COS1 cells transfected with human Shc cDNA (8) using a rabbit antiserum (10) and immobilized onto Protein A-Sepharose. At the same time confluent plates of COS1 cells, transiently expressing epitope (Glu-Glu)-tagged hSOS1, which were either untreated or treated with PMA (1 M) for 30 min, were scraped on ice into 1 ml of 10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 1 mM Pefabloc, 20 g/ml each of leupeptin, aprotinin, soybean trypsin inhibitor, 50 mM NaF, 1 mM NaVO 4 . Following 30 min on ice, cells were disrupted with 40 strokes in a Dounce homogenizer, and the nuclear fraction was removed by centrifugation. The resulting lysate was centrifuged at 100,000 ϫ g for 30 min at 4°C to separate the soluble (S100) fraction from the sediment (P100 fraction) (11). 100 l of the S100 fraction (corresponding to 120 g of total cell protein) were incubated with the Shc-Protein A-Sepharose beads and 3 pmol of GST-Grb2 in 1 ml of lysis buffer. Following 1 h at 4°C, beads were recovered, washed 3 times with 1 ml of lysis buffer. The formation of the hSOS1-GST-Grb2-Shc complex was detected by resolving the immunoprecipitates on SDS-PAGE followed by Western immunoblotting with an anti-Glu-Glu mAb. To ascertain that hSOS1 was expressed at a similar level, 20 l from each S100 fraction were analyzed by Western immunoblotting using an anti-Glu-Glu mAb. In addition, to ascertain that equivalent amounts of Shc immunoprecipitates were used in each condition, one-tenth of the Shc-Protein A-Sepharose beads was retained and analyzed by Western immunoblotting using an anti-Shc mAb (Transduction Laboratories).

RESULTS
hSOS1 Phosphorylation in COS1 Cells-We studied hSOS1 phosphorylation in COS1 cells transiently expressing epitope (Glu-Glu)-tagged hSOS1. Stimulation of the cells with EGF or PMA resulted in hSOS1 phosphorylation, as indicated by its decreased electrophoretic mobility (Fig. 1A). A similar mobility shift was observed in mSOS1 immunoprecipitated from EGFtreated fibroblasts and was reversed by treatment with phosphatases (19). The electrophoretic mobility of hSOS1 was also decreased following co-expression with oncogenic Ras (H-Ras G12V) or with a membrane-targeted, constitutively active ver- FIG. 1. A, hSOS1 phosphorylation in COS1 cells. COS1 cells were transfected with plasmids encoding Glu-Glu (Met-Glu-Tyr-Met-Pro-Met-Glu)-tagged hSOS1 or cotransfected with Glu-Glu hSOS1 and H-Ras G12V, or Raf-CAAX. Following transfection the cells were grown for 3 days, serum starved, and treated with 10 nM EGF for 10 min or 1 M PMA for 30 min as indicated. hSOS1 was immunoprecipitated using a rabbit antiserum raised against the full-length hSOS1 protein and analyzed by Western immunoblotting using an anti-Glu-Glu mAb. B, co-transfection of hSOS1 with dominant negative Ras inhibits EGFinduced phosphorylation of hSOS1. COS1 cells were co-transfected with plasmids encoding Glu-Glu hSOS1 and H-Ras S17N, and stimulated with 3 nM EGF for the indicated times. Glu-Glu hSOS1 was immunoprecipitated using a rabbit antiserum raised against the full-length hSOS1 protein and analyzed by Western immunoblotting using an anti-Glu-Glu mAb.
sion of Raf (Raf-CAAX) (11), suggesting that hSOS1 was phosphorylated in vivo as a result of the activation of the Ras signaling pathway (Fig. 1A). Consistent with these data coexpression of hSOS1 with a dominant negative version of H-Ras, H-Ras S17N, caused a reduction of EGF-induced hSOS1 phosphorylation (Fig. 1B).
GST-Erk1 Phosphorylates hSOS1 in Vitro-hSOS1 contains several potential MAPK phosphorylation sites (Pro-X n -Ser/ Thr-Pro, where X is a neutral or basic amino acid and n ϭ 1 or 2) (31) all situated within the carboxyl-terminal proline-rich domain. We attempted to phosphorylate recombinant hSOS1 in vitro using GST-Erk1 which was pre-activated by incubation with active Raf and Mek in a reconstituted Raf-Mek-MAPK signaling pathway (24). Under these conditions activated GST-Erk1 was able to phosphorylate its substrate myelin basic protein. GST-Erk1 efficiently phosphorylated hSOS1 which showed a decreased electrophoretic mobility and migrated as a single band on SDS-PAGE (not shown). Following a 30-min incubation with GST-Erk1 the stochiometry of hSOS1 phosphorylation was 1.6 mol of phosphate/mol of hSOS1, this ratio did not change when the incubation was prolonged up to 1 h. When GST-Erk1 was omitted from the reaction, neither active Raf nor Mek phosphorylated hSOS1 in vitro. In a parallel set of experiments a commercially available preparation of protein kinase C, including the ␣, ␤, and ␥ isoforms (Boehringer Mannheim), was unable to phosphorylate hSOS1 in vitro. However, the same PKC preparation phosphorylated the PKC substrate Ac-MBP (4 -14) (Boehringer Mannheim) in vitro (not shown).
hSOS1 Phosphorylation Occurs Within the Carboxyl-terminal Proline-rich Region-To map the region of hSOS1 which was phosphorylated in vitro by GST-Erk1, we generated a series of epitope (Glu-Glu)-tagged hSOS1 deletion mutants covering the entire length of the hSOS1 protein ( Fig. 2A). Deletion mutants were transiently expressed in COS1 cells and immunoprecipitated using an anti-Glu-Glu mAb. Immunoprecipitates were phosphorylated in vitro using [␥-32 P]ATP and activated GST-Erk1, then resolved on SDS-PAGE. Gels were dried and subjected to autoradiography which showed that GST-Erk1 phosphorylated only those deletion mutants which included the proline-rich region (amino acids 1066 to 1333) (Fig.  2B). We ascertained that all deletion mutants were expressed at a similar level by analyzing a parallel set of immunoprecipitates by Western immunoblotting (not shown).
Next we studied which region of hSOS1 was phosphorylated in vivo in response to stimulation with EGF. COS1 cells expressing the hSOS1 deletion mutants were metabolically labeled with 32 P i and stimulated with EGF. Deletion mutants were immunoprecipitated and resolved on SDS-PAGE which demonstrated that hSOS1 was phosphorylated in vivo within the region containing amino acids 1066 and 1333, thus within the same region that was phosphorylated in vitro by GST-Erk1 (Fig. 2C). A low amount of hSOS1 phosphorylation was also detected in unstimulated cells (Fig. 2C): this could be due to the high basal level of MAPK activity which has been described in COS1 cells (21). We also studied which region of hSOS1 was phosphorylated in response to stimulation with PMA. These experiments showed that, as with EGF-stimulated cells, treatment with PMA resulted in phosphorylation of hSOS1 only within the proline-rich domain (not shown).
Tryptic Peptide Mapping of Phosphorylated hSOS1-To further characterize the involvement of MAPK in hSOS1 phosphorylation, we carried out two-dimensional tryptic phosphopeptide analysis of hSOS1 phosphorylated in vivo, in EGFstimulated COS1 cells labeled with 32 P i , and in vitro, using GST-Erk1 and [␥-32 P]ATP. The study of the tryptic maps showed that 11 phosphopeptides were generated using in vitro phosphorylated hSOS1 (Fig. 3A) and 13 phosphopeptides were generated using in vivo phosphorylated hSOS1 (Fig. 3B). Analysis of a mixture of radiolabeled phosphopeptides prepared from both conditions showed that 7 of them were generated using either in vitro or in vivo phosphorylated hSOS1 (Fig. 3, A  and B, peptides 1 to 7), strongly suggesting that MAPK phosphorylated hSOS1 in vivo, on sites situated within these peptides. A further 6 phosphopeptides were detected only when analyzing in vivo phosphorylated hSOS1, implying that, besides MAPK, other serine/threonine kinases also phosphoryl- FIG. 2. A, hSOS1 deletion mutants. Dark shaded areas, CDC25 homology domain; light shaded areas, proline-rich region. B, GST-Erk1 phosphorylates hSOS1 within the carboxyl-terminal proline-rich region. Epitope (Glu-Glu)-tagged hSOS1 deletion mutants were transiently expressed in COS1 cells and, following overnight serum deprivation, immunoprecipitated using an anti-Glu-Glu mAb prebound to protein G-Sepharose. Immunoprecipitates were phosphorylated in vitro using [␥-32 P]ATP and GST-Erk1 which had been activated in vitro using a reconstituted Raf-Mek-MAPK signaling pathway. Immunoprecipitates were resolved by SDS-PAGE and phosphorylated proteins were detected by autoradiography. C, hSOS1 is phosphorylated within the proline-rich region in EGF-treated COS1 cells. COS1 cells, transiently expressing epitope (Glu-Glu)-tagged hSOS1 deletion mutants, were metabolically labeled with 32 P i in phosphate-free minimum essential Eagle's medium (0.5 mCi/ml, 2 ml/6 cm dish). At the end of the incubation with the radiolabel, cells were stimulated with 10 nM EGF for 5 min. hSOS1 deletion mutants were immunoprecipitated using an anti-Glu-Glu mAb. Immunoprecipitates were resolved on SDS-PAGE and phosphorylated proteins were detected by autoradiography.
hSOS1 Phosphorylation by MAPK ated hSOS1 in vivo. Of the 11 phosphopeptides generated using in vitro phosphorylated hSOS1, 4 were not found in the tryptic map of in vivo phosphorylated hSOS1. It is possible that kinases other than MAPK phosphorylated these 4 peptides in vivo on additional sites causing a different mobility from that of the corresponding phosphopeptides phosphorylated in vitro by GST-Erk1. Alternatively, the same sites were unavailable to phosphorylation by MAPK in vivo. Treatment of baculovirus produced hSOS1 with phosphatase prior to in vitro phosphorylation did not result in the generation of further phosphopeptide species. Similarly we did not detect changes in the phosphopeptide map of in vivo phosphorylated hSOS1 when we stimulated the cells for a longer time or when we used a lower concentration of EGF.
Phosphorylation by MAPK Does Not Affect hSOS1 Binding to Grb2 in Vitro-The hSOS1 carboxyl-terminal proline-rich region contains the binding sites for the adapter Grb2 (4). We studied if phosphorylation by GST-Erk1 modified the affinity of hSOS1 for GST-Grb2 in vitro. We used either unmodified or in vitro phosphorylated hSOS1 to compete with the binding of 32 P-labeled phosphorylated hSOS1 to GST-Grb2. A the end of a 15-min incubation the hSOS1-GST-Grb2 complexes were captured, resolved on SDS-PAGE, and the amount of 32 P-labeled phosphorylated hSOS1 remaining bound to GST-Grb2 was quantified. Fig. 4 shows that unmodified or phosphorylated hSOS1 demonstrated a nearly identical capacity to compete with 32 P-labeled phosphorylated hSOS1 for binding GST-Grb2 suggesting that the affinity of hSOS1 for Grb2 was not significantly modified by GST-Erk1 phosphorylation.
Phosphorylation by MAPK Decreases hSOS1 Ability to Form a Complex with the hEGFR and GST-Grb2-Proline-rich regions are extended and flexible domains often participating in protein-protein interactions (32). We speculated that phosphorylation of the hSOS1 carboxyl-terminal proline-rich domain regulated, although indirectly, the interaction of hSOS1 with the activated hEGFR. To test this hypothesis, we reconstituted the hSOS1-Grb2-EGFR complex in vitro. We immunoprecipitated the hEGFR from HER14 fibroblasts and incubated the immunoprecipitates with epitope (Glu-Glu)-tagged, recombinant hSOS1 and GST-Grb2. At the end of the 1-h incubation, formation of the hSOS1-GST-Grb2-hEGFR complex was detected by resolving the immunoprecipitates on SDS-PAGE followed by Western immunoblotting using an anti-Glu-Glu mAb. The association of hSOS1 with the hEGFR was entirely dependent on the addition of GST-Grb2 to the reconstitution assay (not shown) and on the activation of the hEGFR, which was achieved by stimulating the HER14 cells with EGF. Furthermore, the amount of hSOS1 that we found associated with the hEGFR immunoprecipitates increased proportionally with the concentration of EGF we used to stimulate the HER14 fibroblasts (Fig. 5A). However, when we used in vitro phosphorylated hSOS1 we found that its ability to form a complex with GST-Grb2 and hEGFR was markedly decreased in each of the conditions we tested (Fig. 5A).
Serine/Threonine Phosphorylation of the hEGFR and hSOS1 Impairs the Formation of the hSOS1-GST-Grb2- The band corresponding to hSOS1 was identified by autoradiography, cut out, and digested with trypsin. Tryptic peptides, corresponding to 10,000 cpm, were applied on cellulose plates, separated by electrophoresis (horizontal) followed by chromatography (vertical), and visualized by autoradiography. The site of application is situated near the bottom right corner of the panel. B, COS1 cells transiently expressing Glu-Glu-tagged hSOS1 were metabolically labeled with 32 P i and stimulated with 10 nM EGF for 2 min. hSOS1 was immunoprecipitated with an anti-Glu-Glu mAb, resolved by SDS-PAGE and analyzed as described in A. AϩB, tryptic phosphopeptides from A and B (corresponding to 7,000 cpm from each sample) were mixed and analyzed as described in A. Seven phosphopeptide species (phosphopeptides 1 to 7) showed identical migration whether they were generated from hSOS1 phosphorylated in vitro or in vivo.

FIG. 4. Phosphorylation by MAPK does not affect hSOS1 binding to GST-Grb2.
Bacterially produced GST-Grb2 (10 pmol) was mixed with 32 P-labeled phosphorylated hSOS1 (6.5 pmol) and the indicated amounts of unmodified hSOS1 (q) or cold-phosphorylated hSOS1 (E). Tubes were rotated at 4°C for 15 min then the GST-Grb2-hSOS1 complexes were captured using glutathione-agarose beads and resolved on SDS-PAGE. The amount of 32 P-labeled phosphorylated hSOS1 bound to GST-Grb2 was quantified using an AMBIS ␤ scanner (100% ϭ 1,000 cpm). Under the conditions of this assay, 32 P-labeled phosphorylated hSOS1 saturated only 50% of GST-Grb2 binding capacity. The averages (ϮS.D.) of three independent experiments are shown.
hEGFR Complex-Serine/threonine phosphorylation of the hEGFR by MAPK and by other kinases such as CaM kinase 2 or protein kinase C also contributes to the acute desensitization of EGF signaling (16,33). To study if serine/threonine phosphorylation of hEGFR affected its capacity to associate with GST-Grb2 and hSOS1, we reconstituted the hSOS1-GST-Grb2-hEGFR complex using hEGFR immunoprecipitated from HER14 fibroblasts which were treated with PMA prior to stimulation with EGF. Treatment with PMA causes phosphorylation of hEGFR on serine and threonine residues which are also phosphorylated in response to prolonged stimulation with EGF and which are important for receptor desensitization (34). Al-though with reduced capacity, unmodified hSOS1 was still able to form a complex with GST-Grb2 and hEGFR immunoprecipitated from PMA-treated cells, whereas in vitro phosphorylated hSOS1 failed to bind through GST-Grb2 to the same hEGFR preparation (Fig. 5B). Thus, serine/threonine phosphorylation of the hEGFR alone was not sufficient to prevent its interaction with hSOS1, whereas phosphorylation of both hSOS1 and hEGFR entirely blocked the assembly and/or impaired the stability of the hSOS1-GST-Grb2-hEGRF complex.
Phosphorylation of the hSOS1 Proline-rich Region Affects Its Interaction with GST-Grb2 and the hEGFR-We used the hSOS1 deletion mutant SOS⌬4, which comprised amino acid residues 1066 to 1333, to study if phosphorylation of the hSOS1 proline-rich region was sufficient to prevent the formation of the hSOS1-GST-Grb2-hEGFR complex. Fig. 5C shows that in vitro phosphorylation by GST-Erk1 decreased the capacity of SOS⌬4 to bind the hEGFR through GST-Grb2.
PMA Treatment Prevents the Interaction of mSOS1 with the hEGFR in Vivo-We used HER14 cells to study the effects of phosphorylation of mSOS1 and hEGFR in vivo. In these cells, a brief treatment with EGF alone induced co-immunoprecipitation of hEGFR and the endogenous mSOS1, which is 98% identical to hSOS1 (1) (Fig. 5D). However, we failed to coimmunoprecipitate mSOS1 and the hEGFR from cells in which serine/threonine phosphorylation of hSOS1 and the hEGFR was induced with PMA prior to EGF stimulation (Fig. 5D). Treatment with PMA inhibits the tyrosine kinase activity of hEGFR and reduces its ability to autophosphorylate on tyrosine residues (34). Nevertheless, our failure to co-immunoprecipitate mSOS1 and hEGFR from PMA-treated cells could be attributed only in part to the diminished tyrosine phosphorylation of the hEGFR, as hEGFR, immunoprecipitated from cells pretreated with PMA and stimulated with EGF, was still able to associate with unmodified hSOS1 in vitro (Fig. 5B).
Phosphorylation of hSOS1 Impairs the Formation of the hSOS1-GST-Grb2-Shc Complex-Another route leading to activation of Ras in response to stimulation with EGF involves the formation of a complex of hSOS1 with tyrosine-phosphorylated Shc and Grb2. We studied if the assembly of the hSOS1-Grb2-Shc complex was also impaired by hSOS1 phosphorylation. To reconstitute the hSOS1-Grb2-Shc complex in vitro, we immunoprecipitated Shc from COS1 cells transfected with Shc cDNA (8). Immunoprecipitates were incubated with GST-Grb2 and with the S100 fraction of a lysate of COS1 cells expressing Glu-Glu-tagged hSOS1, which were either untreated or stimulated with PMA to induce hSOS1 phosphorylation. The formation of the hSOS1-GST-Grb2-Shc complex was detected by resolving the immunoprecipitates on SDS-PAGE followed by Western immunoblotting using an anti-Glu-Glu mAb. Under these conditions hSOS1 formed a complex with GST-Grb2 and Shc and the assembly of such complex was entirely dependent on the addition of GST-Grb2 to the reconstitution assay (Fig.  6A). In contrast, phosphorylated hSOS1 was unable to form a complex with Shc and GST-Grb2 (Fig. 6A).
To establish if phosphorylation of mSOS1 affected its interaction with Shc in vivo, we attempted to co-immunoprecipitate mSOS1 and Shc from HER14 fibroblasts. Although we were able to co-immunoprecipitate Shc and mSOS1 from cells stimulated with EGF, we failed to co-immunoprecipitate Shc and mSOS1 from cells which were pretreated with PMA and stimulated with EGF (Fig. 6B). Treatment of HER14 cells with PMA caused a decrease of Shc tyrosine phosphorylation which could account for our failure of co-immunoprecipitating mSOS1 and Shc. Nevertheless, a decrease of Shc tyrosine phosphorylation was not sufficient to prevent the formation of the hSOS1-GST-Grb2-Shc complex in vitro, as Shc immunoprecipitated 5. A and B, effect of hSOS1 phosphorylation on the assembly of the hSOS1-GST-Grb2-hEGFR complex in vitro. HER14 fibroblasts were serum starved and stimulated for 2 min with the indicated concentrations of EGF, or treated with PMA (1 M) for 30 min prior to EGF stimulation. Cells were lysed and the hEGFR was immunoprecipitated using an anti-hEGFR mAb. Immunoprecipitates were mixed with 3 pmol of recombinant Glu-Glu hSOS1 or in vitro phosphorylated Glu-Glu hSOS1 and with 3 pmol of GST-Grb2. Formation of the hSOS1-GST-Grb2-hEGFR complex was detected by resolving the hEGFR immunoprecipitates on SDS-PAGE followed by Western immunoblotting using an anti-Glu-Glu mAb. C, effect of phosphorylation of hSOS1 prolinerich region (SOS⌬4) on the assembly of the Sos⌬4-GST-Grb2-hEGFR complex in vitro. hSOS1 proline-rich region (amino acid residue 1066 to 1333, Sos⌬4) was expressed in Escherichia coli as a GST fusion protein and epitope (KT3) tagged (40). Unmodified or in vitro phosphorylated SOS⌬4 (3 pmol) was mixed with GST-Grb2 (3 pmol) and hEGFR immunoprecipitates. Association of Sos⌬4 with GST-Grb2 and the hEGFR was detected by Western immunoblotting using an anti-KT3 mAb (40). D, treatment with PMA prevents EGF induced co-immunoprecipitation of mSOS1 and the hEGFR. HER14 cells were serum starved for 16 h and stimulated with the indicated concentrations of EGF for 2 min or treated with PMA (1 M) for 30 min prior to EGF stimulation. Cells were lysed and the hEGFR or mSOS1 was immunoprecipitated using a polyclonal antibody to the hEGFR or to mSOS1 (both from Upstate Biotechnology Inc., Lake Placid, NY). Immunoprecipitates were resolved on SDS-PAGE and analyzed by Western immunoblotting using a mAb to mSOS1 (Transduction Laboratories) or with 2 antibodies to phosphotyrosine (both from Upstate Biotechnology Inc.) which detected a tyrosine phosphorylated protein of approximately 180 kDa corresponding to the activated hEGFR. from PMA-treated cells was still able to bind GST-Grb2 and unmodified hSOS1 (not shown).

DISCUSSION
The formation of complexes of the guanine nucleotide exchange factor hSOS1, the activated hEGFR or Shc, and the adapter Grb2 is critical in coupling EGF stimulation to the activation of Ras. This paper proposes a novel negative feedback mechanism by which the ability of hSOS1 to participate in such signaling complexes is impaired by serine/threonine phosphorylation of its carboxyl-terminal proline-rich domain.
hSOS1 was phosphorylated in vivo as a result of the activation of the Ras signaling pathway by growth factors acting through receptor tyrosine kinases or following co-transfection with activated Ras or Raf. Several lines of evidence suggested that among the serine/threonine kinases acting downstream of Ras, MAPK was a good candidate for phosphorylating hSOS1 in vivo. Blocking EGF-induced MAPK activation by co-transfecting a dominant negative version of Ras resulted in reduction hSOS1 phosphorylation. The carboxyl-terminal proline-rich domain of hSOS1 contains potential MAPK phosphorylation sites and was phosphorylated in vivo, in cells stimulated with EGF. The same proline-rich region of hSOS1 was also phosphorylated in vitro, by activated MAPK. In addition, phosphorylation by MAPK caused a decrease of hSOS1 electrophoretic mobility similar to that of in vivo phosphorylated hSOS1. The study of the two-dimensional tryptic phosphopeptide map confirmed the involvement of MAPK in hSOS1 phosphorylation. Seven phosphopeptides which were detected in the tryptic map of hSOS1 phosphorylated in vitro by GST-Erk1, were also found in the map of hSOS1 phosphorylated in vivo. A further 6 tryptic phosphopeptides were generated only when analyzing hSOS1 phosphorylated in vivo, thereby suggesting that serine/threonine kinases other than MAPK also participated in hSOS1 phosphorylation. Which serine/threonine kinases phosphorylate hSOS1 in addition to MAPK is not yet clear although it is likely that, as MAPK, these enzymes are downstream targets of Ras.
Phosphorylation of hSOS1 occurred within the region containing the binding sites for the SH3 domains of Grb2. Nevertheless, we did not detect any significant difference between the ability of either unmodified or in vitro phosphorylated hSOS1 to bind GST-Grb2. Different from our data, it has recently been suggested that stimulation with insulin of 3T3-L1 adipocites or Chinese hamster ovary cells induced phosphorylation of mSOS1 and disassociation of the mSOS1-Grb2 complex (20). Stimulation with insulin results in the activation of the Ras pathway and of other insulin-specific signaling systems which operate distinct negative feedback mechanisms (35). Therefore it is possible that in response to stimulation with insulin, mSOS1 is phosphorylated by kinases different from those activated by EGF, or that mSOS1-Grb2 interacts with yet unidentified proteins, this resulting in disassociation of the mSOS1-Grb2 complex.
Recombinant hSOS1, phosphorylated using GST-Erk1, showed a decreased capacity to form a complex with GST-Grb2 and the activated hEGFR in vitro. Similarly phosphorylated hSOS1 failed to form a complex with GST-Grb2 and tyrosinephosphorylated Shc in vitro. These data suggest that one effect of MAPK phosphorylation of hSOS1 was to impair its participation in signaling complexes with the hEGFR or Shc, thus uncoupling signal transduction from the hEGFR to Ras. Downregulation of EGF signaling also occurs through the serine/ threonine phosphorylation of the hEGFR which results in inhibition of the tyrosine kinase activity of the receptor and in the decrease of its affinity for EGF (16). In addition, serine/threonine phosphorylation of the hEGFR cooperated with hSOS1 phosphorylation in blocking the assembly of the hSOS1-GST-Grb2-hEGFR complex in vitro.
The mechanism by which phosphorylation of hSOS1 by MAPK interferes with the participation of hSOS1 in signaling complexes is not clear. It is possible that phosphorylation by MAPK decreases the stability and induces the disassembly of the hSOS1-GST-Grb2-hEGFR complex. It is also possible that, binding of Grb2 to the hyperphosphorylated carboxyl terminus of hSOS1 results in alteration of Grb2 structure and loss of its affinity for binding tyrosine-phosphorylated proteins. The study of the crystal structure of Grb2 suggests that a conformational change affecting the SH3 domains of Grb2 would be unlikely to alter the SH2 domain (36). Nevertheless, binding of the SH2 domain of Grb2 to tyrosine-phosphorylated Shc positively regulates the interaction of the SH3 domains with the proline-rich region of mSOS1 (37). Thus, conformational changes appear to be transmitted through Grb2 and these changes are involved in regulating Grb2 interaction with target proteins.
MAPK is not the only kinase that phosphorylates hSOS1 in vivo, therefore, to fully understand the biochemical function of hSOS1 phosphorylation, it will be necessary to identify the other serine/threonine kinases as well as the sites that are phosphorylated. It is plausible that phosphorylation of hSOS1 in vivo, by analogy with the effects of phosphorylation by MAPK in vitro, also impairs the participation of hSOS1 in complexes with hEGFR or Shc. Indeed, EGF failed to induce co-immunoprecipitation of mSOS1 and hEGFR or Shc in cells which were pretreated with PMA. In conclusion our data suggest that hSOS1 phosphorylation, by limiting the access of hSOS1 to Ras, contributes to regulate the duration or the amplification of signals originating from activated tyrosine kinases. In agreement with this view, prolonged Ras⅐GTP loading has been detected in insulin-stimulated cells in which phosphorylation of mSOS1 was inhibited using a specific Mek inhibitor (38). It has been proposed that the duration of the activation of the Ras pathway (sustained versus transient) is critical to enact FIG. 6. Effect of hSOS1 phosphorylation on the formation of the hSOS1-Grb2-Shc complex. A, COS1 cells were transfected with Shc cDNA and stimulated with the indicated concentrations of EGF. Shc was immunoprecipitated using a rabbit antiserum raised against the full-length Shc protein. The immunoprecipitates, immobilized on Protein A-Sepharose beads, were mixed with GST-Grb2 (3 pmol) and with 100 l of a soluble fraction (S100) of COS1 cells expressing Glu-Glu hSOS1 which were either untreated or treated with PMA (1 M) for 30 min, to induce hSOS1 phosphorylation. The formation of the Shc-GST-Grb2-hSOS1 complex was detected by analyzing the immunoprecipitates on SDS-PAGE followed by Western immunoblotting with an anti-Glu-Glu mAb. B, HER14 cells were stimulated with the indicated concentrations of EGF for 2 min or treated with PMA (1 M) for 30 min prior to EGF stimulation. Cells were lysed and Shc was immunoprecipitated using a rabbit antiserum raised against the full-length Shc protein. Immunoprecipitates were analyzed by Western immunoblotting using a mAb to mSOS1. specific cellular responses (39). According to this model, hSOS1 phosphorylation contributes to secure the execution of the appropriate cellular programs in response to growth factor stimulation.