ERK Regulates the Hepatocyte Growth Factor-mediated Interaction of Gab1 and the Phosphatidylinositol 3-Kinase*

Based on our previous observations that active ERK associates with and phosphorylates Gab1 in response to HGF, and the prediction that the ERK phosphorylation site is adjacent to one of the phosphatidylinositol 3-ki-nase (PI3K) SH2 binding motifs, we examined the possibility that ERK phosphorylation can regulate the Gab1/ PI3K association. The HGF-mediated association of Gab1 with either full-length GST-p85 or its isolated N- or C-terminal SH2 domains was inhibited by (cid:1) 50% in the setting of ERK inhibition, a result confirmed by co-immunoprecipitation of the native proteins. A 14-amino acid peptide encoding 472 YVPMTP 477 (one of the major p85 binding sites in Gab1 and the predicted ERK phosphorylation site) was synthesized with either phosphotyrosine alone (pY), or phosphotyrosine (cid:1) phosphothreonine (pYT). In both pull-down assays and competition assays, pYT demonstrated a higher affinity for p85 than did pY alone. Finally, examination of the phosphorylation state of Akt after HGF stimulation revealed that ERK inhibition resulted in a decrease in Akt activation at both 5 and 10 min. These results sugg-est that activated ERK can phosphorylate Gab1 in response The effects of HGF on c-Met activation and subsequent activation of the PI3K has been extensively studied in these cells Ex- periments performed confluent. Growth Factor Stimulation and MEK Inhibition— mIMCD-3 cells were serum-starved for 24 h in Dulbecco’s modified Eagle’s medium/ F-12 and then stimulated with HGF (40 ng/ml) or vehicle control for 20 min. To inhibit ERK activation, cells were pretreated with 10 (cid:2) M U0126 (Promega) for 20 min prior to HGF stimulation. This dose was chosen following a dose-response curve in these cells in which 1 (cid:2) M U0126 was found to inhibit HGF-dependent ERK activation by 90%, whereas 10 (cid:2) M U0126 inhibited ERK activation completely (22). At doses less than 50 (cid:2) M , U0126 has been found to be selective for the MEK family of dual- specificity kinases (23), with inhibition of MEK1, MEK2, and MEK5. Following HGF stimulation, cells were lysed in 800 (cid:2) l of ice-cold lysis buffer containing 50 m M Tris, pH 7.5, 150 m M NaCl, 1 m M EDTA, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 2 (cid:2) g/ml aprotinin, 2 (cid:2) g/ml leupeptin, 2 (cid:2) g/ml pepstatin A, 10 m M NaF, 1 m M phenylmeth-ylsulfonyl fluoride, and 1 m M Na 3 VO 4 . GST Fusion Protein Expression and Pull-down— pGEX plasmids encoding full-length p85, the carboxy SH2 domain (CSH2), and the more N-terminal SH2 domain (NSH2) as GST fusion proteins were kindly provided by Dr. Lucia Rameh (Harvard University). Expression of the fusion proteins was induced in BL21-competent bacteria with isopropyl-(cid:3) - D -thiogalactoside as previously described (18). Briefly, the bacteria were lysed with sodium deoxycholate, and the supernatant was col-lected and incubated with a 50% slurry of glutathione-Sepharose 4B (Amersham Pharmacia Biotech). The beads were washed and resus-pended in 50 m M HEPES, pH 7.4, 150 m M NaCl, 5 m M dithiothreitol, 5% glycerol, and 1 (cid:2) g/ml leupeptin until use. Control experiments were performed with GST-Sepharose beads generated by expression of GST alone, using the empty pGEX-4T vector. Total GST fusion protein amounts were estimated virtually using Coomassie Blue-stained SDS- PAGE with albumin standards. One mg of whole cell lysate protein containing (cid:1) 0.1 (cid:2) g of endogenous p85 was then incubated with 50 (cid:2) g of the GST fusion protein of interest for 30 min at 4 °C. The glutathione beads were washed three times with ice-cold lysis buffer followed by resolution via SDS-polyacrylamide gel electrophoresis. Proteins were electrophoretically transferred onto Immobilon-P transfer membranes (Millipore) using a Trans-Blot SD Semi-Dry Transfer Cell (Bio-Rad) for 90 min at 18 V. Usually, membranes were blocked fo r 1 h atroom temperature with 5% nonfat dry milk in wash buffer containing 10 m M Tris, pH 7.5, 100 m M NaCl, and 0.1% Tween-20 (TBS-T). After five additional rinses with TBS-T, blots were incubated with primary antibody to p85 (Upstate Biotechnology Inc.). After rinsing with TBS-T, membranes were incubated with horseradish peroxidase-conjugated secondary antibody at 1:5000 dilution in TBS-T for 60 min at room temperature. Blots were visualized by an ECL system (Amersham Pharmacia Biotech). Co-immunoprecipitation and Western Blotting— 600 (cid:2) g of mIMCD-3 cell lysates were pre-cleared fo r 1 h at 4°C with protein A-Sepharose CL 4B (1:1

display developmental defects in myotomes, placenta, and liver that recapitulate the loss of HGF and c-Met (8).
The phenotypic effects of Gab1 signaling have been found to be critically dependent on Gab1 interactions with the PI3K. The PI3K is a heterodimer composed of an 85-kDa regulatory subunit (p85) and a 110-kDa catalytic subunit (p110). Two SH2 domains are located in the C-terminal region of the p85 subunit and have been shown to bind to membrane-associated tyrosinephosphorylated proteins, resulting in recruitment of the PI3K to the membrane and activation of the lipid kinase activity of the 110-kDa subunit. This activity results in the formation of phosphoinositide 3,4,5-trisphosphate (PI 3,4,5 P 3 ) at the membrane, which in turn serves to recruit and help activate such proteins as protein kinase C (PKC) and Akt (9,10). The activated PI3K has been demonstrated to result in the regulation of various cellular activities including proliferation (11), differentiation (12), and prevention of apoptosis (13). We have found that PI3K activity is critical for c-Met-mediated cell migration and in vitro tubulogenesis (14), at least in part by activating PKC (15). Another role for PI3K in morphogenic signaling has been demonstrated by the observation that the PH domain of Gab1 binds PI 3,4,5 P 3 , which serves to recruit Gab1 to the membrane following PI3K activation (1,16,17). Loss of the PH domain of Gab1 results in loss of epithelial morphogenesis following HGF stimulation, despite recruitment of Gab1 to the c-Met receptor and its phosphorylation (16).
We have recently demonstrated that in addition to SH2 and SH3 domain containing proteins, Gab1 also interacts with phosphorylated ERK1 and 2 following HGF or EGF stimulation of epithelial cells (18). The association of ERK with Gab1 results in the phosphorylation of Gab1 on serine and threonine residues, primarily in the Met binding domain (MBD). This domain also includes one of three major consensus sequences for PI3K binding ( 472 YVPM 475 ), as well as a predicted ERK phosphorylation site immediately adjacent ( 472 YVPMTP 477 ). We therefore hypothesized that ERK phosphorylation of this site in vivo might result in secondary regulation of p85 SH2 domain binding to Gab1. In the present study, we confirm that HGF stimulation results in Gab1-p85 association, and demonstrate that this interaction is partly dependent on ERK activation. Furthermore, we show that inhibition of ERK activation not only decreases the association of p85 with Gab1 but also decreases the downstream phosphorylation of Akt.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-Immortalized mIMCD-3 epithelial cells (19) were maintained using standard cell culture techniques in Dulbecco's modified Eagle's medium/F-12 containing 10% fetal bovine serum. The effects of HGF on c-Met activation and subsequent activation of the PI3K has been extensively studied in these cells (14,15,20,21). Experiments were performed when the cells were 80 -90% confluent. All reagents were obtained from Sigma Chemical Co. unless otherwise noted.
Growth Factor Stimulation and MEK Inhibition-mIMCD-3 cells were serum-starved for 24 h in Dulbecco's modified Eagle's medium/ F-12 and then stimulated with HGF (40 ng/ml) or vehicle control for 20 min. To inhibit ERK activation, cells were pretreated with 10 M U0126 (Promega) for 20 min prior to HGF stimulation. This dose was chosen following a dose-response curve in these cells in which 1 M U0126 was found to inhibit HGF-dependent ERK activation by 90%, whereas 10 M U0126 inhibited ERK activation completely (22). At doses less than 50 M, U0126 has been found to be selective for the MEK family of dualspecificity kinases (23), with inhibition of MEK1, MEK2, and MEK5. Following HGF stimulation, cells were lysed in 800 l of ice-cold lysis buffer containing 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 2 g/ml aprotinin, 2 g/ml leupeptin, 2 g/ml pepstatin A, 10 mM NaF, 1 mM phenylmethylsulfonyl fluoride, and 1 mM Na 3 VO 4 .
GST Fusion Protein Expression and Pull-down-pGEX plasmids encoding full-length p85, the carboxy SH2 domain (CSH2), and the more N-terminal SH2 domain (NSH2) as GST fusion proteins were kindly provided by Dr. Lucia Rameh (Harvard University). Expression of the fusion proteins was induced in BL21-competent bacteria with isopropyl-␤-D-thiogalactoside as previously described (18). Briefly, the bacteria were lysed with sodium deoxycholate, and the supernatant was collected and incubated with a 50% slurry of glutathione-Sepharose 4B (Amersham Pharmacia Biotech). The beads were washed and resuspended in 50 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM dithiothreitol, 5% glycerol, and 1 g/ml leupeptin until use. Control experiments were performed with GST-Sepharose beads generated by expression of GST alone, using the empty pGEX-4T vector. Total GST fusion protein amounts were estimated virtually using Coomassie Blue-stained SDS-PAGE with albumin standards. One mg of whole cell lysate protein containing ϳ0.1 g of endogenous p85 was then incubated with 50 g of the GST fusion protein of interest for 30 min at 4°C. The glutathione beads were washed three times with ice-cold lysis buffer followed by resolution via SDS-polyacrylamide gel electrophoresis. Proteins were electrophoretically transferred onto Immobilon-P transfer membranes (Millipore) using a Trans-Blot SD Semi-Dry Transfer Cell (Bio-Rad) for 90 min at 18 V. Usually, membranes were blocked for 1 h at room temperature with 5% nonfat dry milk in wash buffer containing 10 mM Tris, pH 7.5, 100 mM NaCl, and 0.1% Tween-20 (TBS-T). After five additional rinses with TBS-T, blots were incubated with primary antibody to p85 (Upstate Biotechnology Inc.). After rinsing with TBS-T, membranes were incubated with horseradish peroxidase-conjugated secondary antibody at 1:5000 dilution in TBS-T for 60 min at room temperature. Blots were visualized by an ECL system (Amersham Pharmacia Biotech).
Co-immunoprecipitation and Western Blotting-600 g of mIMCD-3 cell lysates were pre-cleared for 1 h at 4°C with protein A-Sepharose CL 4B (1:1 slurry in phosphate-buffered saline (Amersham Pharmacia Biotech) and centrifuged at 5000 rpm for 2 min at 4°C. For Gab1 overexpression experiments, lysates from mIMCD-3 cells transiently transfected with FLAG-Gab1 or vector control were utilized. Transfections were performed with LipofectAMINE 2000 (Life Technologies) as previously described (18). Supernatants were incubated with the appropriate antibody (anti-Gab1 or anti-p85 (Upstate Biotechnology Inc.)) overnight followed by addition of protein A-Sepharose CL4B. After incubating for another 2 h at 4°C, samples were centrifuged at 5000 rpm for 2 min at 4°C and washed three times with 0.5 ml of lysis buffer prior to resolution via SDS-PAGE. Western blot was performed using the methods mentioned above or following the protocol of the primary antibody manufacture.
Peptide Synthesis and p85 Pull-down-Three 14-amino acid peptides corresponding to the Gab1 amino acid sequence 466 -479 were synthesized at the Tufts University Core Facility (Boston, MA). The peptide sequences are: PIQEPNYVPMTPGT (Y); PIQEPNpYVPMTPGT (pY); and PIQEPNpYVPMpTPGT (pYT). The N-terminal of these peptides was conjugated to biotin and 30 nmol of each peptide was bound to 1 ml of SoftLink Soft Release Avidin Resin (Promega). The affinity of these peptides for p85 was examined by incubating 1 mg of mIMCD-3 whole cell lysate with resin-conjugated peptide at the indicated concentration for 60 min, followed by washing three times in lysis buffer and separation by SDS-PAGE. Quantitation of p85 was performed by pooling the data from three to five peptide pull-down experiments and normalizing to the amount of p85 binding to pY peptide at 1 pM using the National Institutes of Health IMAGE program.
For competition experiments, we utilized the bacterial co-expression of the inter-kinase PDGF receptor p85 binding domain and the phosphotyrosine kinase Elk to generate a phosphorylated GST-PDGFR fusion protein with a high affinity for p85. These constructs were kindly provided by Saskia Brackmann (Harvard University) and the GST fusion protein was expressed as previously described (see above (24)). One mg of mIMCD-3 whole cell lysates was incubated with the appropriate concentration of non-conjugated synthetic peptide (pY or pYT) in solution at 4°C for 60 min, followed by incubation of the mixture for 15 min with the GST-PDGF receptor fusion protein on glutathione beads. The beads were then washed three times with lysis buffer to remove p85 that was not bound to the GST-PDGFR, and the bound p85 was detected by SDS-PAGE followed by Western blotting.
Statistical Analysis-All experiments were repeated on at least three separate occasions. Quantification of ECL immunoblots was performed using the IMAGE program. Values are expressed as means Ϯ S.E. Results were analyzed using the Student t test. A value of p Ͻ 0.05 was considered significant.

Gab1 Is Phosphorylated by Activated ERK in Response to c-Met Engagement-
We have previously shown that Gab1 acts as a substrate for ERK2 in vitro and following overexpression of Gab1 and constitutively active MEK1 in HEK cells (18). ERK phosphorylation of Gab1 was determined by both gel retardation of the phosphorylated protein as well as peptide mapping.
To determine if growth factor-induced ERK activation could mediate Gab1 phosphorylation, we examined the gel mobility of Gab1 following stimulation with HGF. Prolonged electrophoresis of unstimulated mIMCD-3 cells revealed that Gab1 runs as a doublet at ϳ110 kDa, with the majority of the protein Immunoblotting with Gab1 reveals that HGF induces the association of Gab1 with p85, both via the N-and C-terminal SH2 domains. Pretreatment with U0126 markedly decreases this association. seen at the lower molecular mass (Fig. 1, lane 1). When cells were pre-treated with the MEK inhibitor U0126, the upper band of this doublet disappeared and all of the protein was found at the lower apparent molecular mass (Fig. 1, lane 2). In contrast, following HGF stimulation there is an upward shift in apparent molecular weight with the majority of Gab1 found at the higher molecular weight at both 10 and 20 min of HGF stimulation. This decreased gel mobility was prevented in the presence of U0126. Blotting of these same lysates with an antibody to phosphorylated ERK confirms that the HGF-dependent ERK activation is inhibited by U0126 (Fig. 1, middle  panel). The U0126-inhibitable change in apparent molecular weight of Gab1 is typical of that seen following phosphorylation of proteins at a serine/threonine-proline motif (25) and is indistinguishable from that seen with in vitro phosphorylation of Gab1 by purified active ERK (18). These results therefore suggest that ERK phosphorylates Gab1 in intact renal epithelial cells following HGF stimulation.
ERK Phosphorylation Regulates the Interaction of Gab1 and the PI3K-Based on the juxtaposition of the predicted ERK phosphorylation site in Gab1 with a p85 SH2 domain association site, we predicted that ERK phosphorylation might regulate p85 binding to Gab1. To examine this, whole cell lysates from control and HGF-stimulated mIMCD-3 cells were incubated with glutathione beads linked to GST-p85, and precipitants were examined for the presence of Gab1 (Fig. 2). In HGF-stimulated lysates Gab1 was found to co-precipitate with GST-p85, confirming previous studies that demonstrated the specific nature of this interaction ( Fig. 2A). However, HGFstimulated lysates in which ERK activation had been inhibited by pre-treatment with U0126 revealed a marked reduction in the Gab1 association with p85. To determine if this effect was specific for the p85 SH2 domain interactions with Gab1, we utilized GST fusion proteins encoding the N-terminal and Cterminal SH2 domains of p85. These pull-down experiments revealed that both the N-terminal and C-terminal p85 SH2 domains are capable of associating with Gab1 in an HGF-dependent manner, and further demonstrated that these SH2 domain interactions are partially dependent on MAPK activation (Fig. 2B).
The Gab1-PI3K Interaction in Intact Cells Is Partially ERKdependent-To investigate the role of MAPK phosphorylation on the interaction between Gab1 and PI3Kin intact cells, we transiently transfected FLAG-tagged Gab1 into mIMCD-3 cells and examined the interaction between FLAG-Gab1 and p85. We found that FLAG-Gab1 interacted with p85 upon HGF stimulation and that this association was also substantially inhibited by pre-treatment with U0126 (Fig. 3A). We next examined the interaction of native Gab1 with native p85 by co-immunoprecipitation. In both anti-Gab1 immunoprecipitates and anti-p85 immunoprecipitates, we detected association of Gab1 and p85 following stimulation of cells with HGF ( Fig. 3, B and C). Again, following inhibition of MAPK activation with U0126, there was a substantial reduction in co-immunoprecipitation of Gab1 and p85. Quantitation of the effect of MAPK inhibition on Gab1-p85 association was performed by pooling the data from three anti-Gab1 immunoprecipitation experiments and normalizing to the amount of p85 immunoprecipitated following HGF stimulation in the absence of U0126. This revealed a 50% decrease in the association of Gab1 and p85 in the setting of MEK inhibition (Fig. 3D, n ϭ 3, p Ͻ 0.01). These results confirm that p85 associates with Gab1 in an HGF-dependent manner, presumably due to tyrosine phosphorylation of Gab1 by c-Met, and further demonstrates that the affinity of Gab1 for p85 appears to be additionally modulated by ERK activation.

ERK Inhibition Does Not Appear to Alter c-Met-Gab1
Interactions-Based on our prior work demonstrating that ERK directly phosphorylates Gab1, we predicted that the decrease in p85-Gab1 interaction following U0126 treatment was due to a direct effect of phosphorylation of Gab1 near the p85 SH2 binding site. However, nonspecific effects of U0126 on the ability of HGF to induce c-Met activation and/or c-Met-Gab1 association and phosphorylation were also considered. To determine if treatment with U0126 altered c-Met phosphorylation following HGF stimulation, whole cell lysates from cells Ϯ HGF were immunoprecipitated with anti-phosphotyrosine and c-Met was detected by immunoblotting. Upon HGF stimulation, c-Met was tyrosine phosphorylated equally in the presence and absence of pre-treatment with U0126 (Fig. 4A), dem-FIG. 3. The HGF-stimulated Gab1-p85 association in intact cells is partially ERK-dependent. A, either FLAGtagged full-length Gab1 (Flag-Gab1) or empty pFLAG-CMV-II vector (V) was transiently transfected into mIMCD-3 cells followed by HGF (40 ng/ml) stimulation for 20 min Ϯ U0126 (10 M) pretreatment. Anti-FLAG immunoprecipitation and anti-p85 immunoblotting reveal that HGF induces a marked increase in Gab1-p85 association that is partially inhibited by U0126 (upper panel). Immunoprecipitation of either endogenous p85 (B) or Gab1 (C) from untransfected cells stimulated with HGF Ϯ pretreatment with U0126 confirms the marked increase in the co-immunoprecipitation of the two proteins following HGF stimulation and again reveals a substantial inhibition of this increase in the setting of ERK inhibition. Quantitation of the co-immunoprecipitation (D), normalized to 1 for HGF stimulation alone, reveals a 50% decrease in the co-immunoprecipitation of Gab1 and p85 following U0126 treatment (*p Ͻ 0.01, n ϭ 3).
onstrating that MEK inhibition did not detectably alter c-Met activation by HGF.
Following the tyrosine phosphorylation of c-Met, association with Gab1 occurs via both a direct interaction between the MBD of Gab1 and tyrosine 1349 of c-Met, as well as an indirect interaction involving Grb2 binding to tyrosine 1356 of c-Met via its SH2 domain and to a proline-rich region in the MBD of Gab1 via its SH3 domain (2,3,26). Thus, phosphorylation of Gab1 by ERK could regulate the interaction of Gab1 and Grb2 and thereby affect the c-Met-Gab1 association and subsequent Gab1 phosphorylation. To examine this possibility, Gab1 was immunoprecipitated from control and HGF-stimulated mIMCD-3 cell lysates Ϯ pre-treatment with U0126 and co-immunoprecipitation of Grb2 examined (Fig. 4B). As previously described, the association of Gab1 and Grb2 was found to be constitutive and was not altered by treatment with HGF or by MEK inhibition. To more carefully determine if MEK inhibition altered the ability of c-Met to associate with and phosphorylate Gab1, we examined the tyrosine phosphorylation state of Gab1 following HGF stimulation (Fig. 4C). Immunoprecipitates of Gab1 following HGF stimulation revealed a marked increase in Gab1 tyrosine phosphorylation that was not altered by pretreatment with U0126. Taken together, these results demonstrate that pre-treatment with the MEK inhibitor U0126 did not detectably alter the activation of c-Met or its interaction with and phosphorylation of Gab1. Of note, the phosphotyrosine blot of Gab1 is probably not sensitive enough to detect a change in the phosphorylation state at a single residue (see "Discussion").
Phosphorylation at Thr 476 of Gab1 Can Regulate the Gab1/ PI3K Interaction-As previously noted, 472 YVPMTP 477 in the MBD of Gab1 is predicted to be both a major site for PI3K binding as well as an ERK phosphorylation site (27). Based on the decreased affinity of Gab1 for p85 following inhibition of ERK activation, we predicted that dual phosphorylation of this motif might result in a higher affinity for p85 binding than tyrosine phosphorylation alone. To examine this hypothesis, we synthesized biotin-linked 14-amino acid peptides flanking this motif in which either Tyr 472 (pY) or both Tyr 472 and Thr 476 (pYT) were phosphoamino acids, as well as the control peptide in which neither amino acid was phosphorylated (Y). These peptides were linked to streptavidin beads and incubated with 1 mg of mIMCD-3 cell lysates predicted to contain ϳ1 pM of p85. 2 A dose-response curve for each peptide revealed that at peptide concentrations between 0.5 and 2 pM, pYT demonstrated a higher affinity for p85 than did pY alone (Fig. 5, A and  B). In contrast, the Y peptide failed to precipitate p85 at any concentration tested. At concentrations of Ն6 pM, both the pY and pYT peptide pulled down essentially all of the p85 from the supernatant.
To further examine the effect of dual phosphorylation of the peptide on the affinity for p85, we performed competition experiments using a GST-PDGF receptor fusion protein containing the p85 binding domain of the PDGFR phosphorylated by co-expression of the Elk tyrosine kinase (24). Each peptide (200 nM) was incubated with 1 mg of whole cell lysate for 1 h, followed by addition of 10 g of the GST-PDGF receptor conjugated to glutathione beads. p85 that competed off of the phosphopeptide and associated with the GST-PDGFR fusion protein was then detected by precipitation and SDS-PAGE. Consistently, incubation of lysates with pYT resulted in a greater competition for p85 binding than did incubation with the pY peptide (Fig. 6A, compare lanes 2 and 3). In the absence of competitive peptide, GST-PDGFR precipitated essentially all of the p85 in the cell lysates (Fig. 6A, lane 1). Quantitative analysis revealed that incubation with pY resulted in a 60% reduction in GST-PDGFR p85 pull-down, whereas incubation with pYT resulted in a 77% reduction (Fig. 6B).
HGF-mediated PKB/Akt Activation Is Partially Inhibited by Blocking ERK Activation-The activation of PKB/Akt is dependent on membrane localization and activation of the PI3K with subsequent PI 3,4,5 P 3 production (28). Because inhibition of ERK activation decreased HGF-mediated Gab1/PI3K association, we postulated that membrane recruitment of the PI3K would be diminished and subsequent Akt activation decreased. To test this we examined the phosphorylation state of PKB/Akt in IMCD cells stimulated with HGF Ϯ U0126. Treatment with 2 L. Cantley, personal communication.

FIG. 4. ERK inhibition does not alter activation of c-Met, Grb2/ Gab1 association, or Gab1 tyrosine phosphorylation.
A, antiphosphotyrosine immunoprecipitation of whole cell lysates from control and HGF-stimulated mIMCD-3 cells Ϯ U0126 was performed, followed by immunoblotting with anti-c-Met. HGF induces c-Met tyrosine phosphorylation, with no detectable difference in the setting of U0126 treatment. Equality of loading was judged by blotting 20 g of the original lysate with anti-c-Met (lower panel). B, the association of Gab1 with Grb2 was examined following HGF stimulation Ϯ U0126 treatment by co-immunoprecipitation. As previously demonstrated, Gab1 and Grb2 are constitutively associated with no change following HGF stimulation or ERK inhibition (upper panel). The marked size discrepancy between Gab1 and Grb2 prevented transfer of immunoprecipitated Gab1 onto the membrane, therefore equal loading was determined by immunoblotting 20 g of the original cell lysate with anti-Gab1 (lower panel). C, tyrosine phosphorylation of Gab1 was examined by anti-Gab1 immunoprecipitation and anti-pTyr immunoblotting. HGF stimulation results in a marked increase in tyrosine phosphorylation of Gab1, with no change in the setting of ERK inhibition (upper panel). Stripping of the membrane and re-probing with anti-Gab1 demonstrates equal immunoprecipitation of Gab1 (lower panel).
HGF resulted in a detectable increase in Akt phosphorylation by 2 min, with a further increase at 5 and 10 min (Fig. 7). In the setting of pretreatment with U0126, the initial activation of Akt 2 min after HGF stimulation was unchanged, but the further increases seen at 5 and 10 min were prevented. DISCUSSION In the present study we demonstrate that the addition of HGF to cells expressing the c-Met receptor results in both tyrosine phosphorylation of Gab1 as well as a decrease in the gel mobility of Gab1. The appearance of a slower mobility on SDS gels is typical of proteins that undergo phosphorylation on serine or threonine residues, particularly in the motif SP or TP. This is apparently because phosphorylation at these sites causes a change in the tertiary structure of the protein by creating a binding site for peptidyl-prolyl cis/trans isomerases such as Pin1 (25,29). In our previous study, we had found that the serine/threonine kinase ERK interacted with the MBD of Gab1 directly and could phosphorylate Gab l in vitro and in vivo resulting in an indistinguishable reduction in gel mobility (18). Consistent with this data, we now show that inhibition of ERK activation prevents the HGF-mediated decrease in Gab1 gel mobility, arguing that ERK may directly phosphorylate Gab1 following HGF stimulation.
ERK typically phosphorylates proteins at the motif PX(S/T)P (30, 31), often followed closely by an ERK binding motif consisting of FXFP (32). Examining the sequence of Gab1 using the ScanSite program reveals one site in the Gab1 sequence that closely fits this preferred motif, 472 YVPMTPGTFDFS 484 . This site is in the Met binding domain (MBD) of Gab1, an 82-amino acid region of the protein that mediates the interaction of c-Met and Gab1 (2), which we have shown to selectively mediate ERK association and to be phosphorylated preferentially by ERK in vitro (18). Interestingly, this ERK phosphorylation site is embedded in the classic PI3K SH2 association site, YVPM, and mutation of the tyrosine residue in this site has been shown to markedly decrease PI3K binding (26,33). This suggests that ERK phosphorylation of this site might regulate the interaction of Gab1 and the PI3K.
To test the role of endogenous ERK in the Gab1-PI3K association, we examined the ability of the p85 subunit of the PI3K to associate with Gab1 following HGF stimulation in the presence or absence of ERK inhibition. As expected, stimulation FIG. 5. Affinity of synthetic peptides for the p85 subunit of the PI3K. Three peptides corresponding to the 14-amino acid PI3K binding site in Gab1 were synthesized (either unphosphorylated (Y), tyrosinephosphorylated (pY), or tyrosine-and threonine-phosphorylated (pYT)) and used to pull down p85 from mIMCD-3 cell lysates. A, peptide Y failed to pull down p85 at any concentration tested, whereas both pY and pYT were capable of associating with p85. At lower concentrations, pYT demonstrated greater pull-down of p85 than did pY. B, quantitation of p85 pull-down reveals that, between 0.5 and 2.0 pM, pYT brings down more p85 than does pY (compared with pY, the relative pull-down of p85 for pYT was 3-fold, 1.7-fold, and 1.5-fold at 0.5 pM (p Ͻ 0.05, n ϭ 3), 1 pM (p Ͻ 0.001, n ϭ 5), and 2 pM (p Ͻ 0.05,n ϭ 3), respectively). At concentrations of Ͼ6 pM there was no difference in p85 pull-down by the two peptides. Data were normalized to 1 for p85 pull-down by the pY peptide at 1 pM.
FIG. 6. Peptide pYT is more efficient than pY in competing for p85 binding. A, pY or pYT peptides (200 nM) that had not been streptavidin-conjugated were incubated for 1 h with 1 mg of mIMCD-3 whole cell lysates, followed by addition of tyrosine-phosphorylated GST-PDGF receptor fusion protein on glutathione beads for 15 min. The beads were then collected, and PDGFR-associated p85 was detected by immunoblotting. Both peptides competed for binding of p85 to the PDGFR construct, with pYT demonstrating a higher affinity for p85. B, quantitative analysis of these competition assays reveals that pY decreased the association of p85 with the GST-PDGFR fusion protein by 60% whereas pYT decreased the association by 77% (n ϭ 3; *, p ϭ 0.05 for pY versus pYT). Data were normalized to 1 for p85 pull-down by the GST-PDGFR in the absence of competitive peptide.

FIG. 7. HGF-mediated PKB/Akt activation is diminished by inhibition of ERK activation.
Serum-starved mIMCD-3 cells were stimulated with HGF (40 ng/ml) for the indicated time Ϯ pre-treatment for 20 min with U0126 (10 M). Whole cell lysates (20 g) were separated by SDS-PAGE and were analyzed by anti-phospho-Akt (upper panel), anti-phospho-ERK (middle panel), and anti-total Akt (lower panel). Akt is activated within 2 min following HGF stimulation, with a further increase in phosphorylation at 5 and 10 min. Pretreatment with U0126 did not alter the initial increase in Akt phosphorylation, but prevented the further rise at 5 and 10 min.
with HGF resulted in a marked increase in the affinity of p85 for Gab1, both in GST-p85 pull-down experiments and in coimmunoprecipitation of the native proteins. This increased affinity of Gab1 for p85 was reproduced equally for both the Nand C-terminal SH2 domains of p85, demonstrating its requirement for tyrosine phosphorylation of the Gab1 protein. However, pretreatment with the MEK inhibitor U0126 resulted in a substantial decrease in the p85-Gab1 interaction in both the pull-down and co-immunoprecipitation experiments, suggesting that ERK phosphorylation of Gab1 increases the affinity of Gab1 for p85. This result was unexpected, because ERK-mediated serine phosphorylation of the structurally similar docking protein, IRS-1, was found to have an opposite effect (34).
In the case of IRS-1, phosphorylation by several serine/threonine kinases, including ERK, PKC, and Akt, has been shown to decrease the affinity of IRS-1 for p85, at least in some cases by decreasing tyrosine phosphorylation of IRS-1 (34 -36). Because the predicted ERK phosphorylation site in Gab1 is in the Met binding domain, we examined the possibility that inhibition of ERK activation was altering the association of c-Met and Gab1 and subsequently the tyrosine phosphorylation state of Gab1. We did not detect a difference in the association between Gab1 and its c-Met adapter protein Grb2, nor did we find a change in the total tyrosine phosphorylation state of Gab1 following HGF stimulation in the presence or absence of MEK inhibition. Thus it does not appear that treatment with U0126 prevents HGF-mediated c-Met activation or c-Met-dependent tyrosine phosphorylation of Gab1. Of note, because Gab1 is known to be potentially phosphorylated on at least 8 tyrosine residues (37), a selective loss of tyrosine phosphorylation of one of the residues critical for p85 association might not be detected by this method.
To determine whether phosphorylation of threonine 476 (the predicted ERK phosphorylation site in Gab1) alters the affinity of tyrosine 472 (the most important residue for p85 binding to Gab1 (33)) for binding the SH2 domain of p85, three peptides based on the 14 amino acids encompassing this site were synthesized. As expected, the non-phosphorylated peptide failed to interact with p85, whereas pY 472 pulled-down p85 at concentrations as low as 1 pM. However, in both simple pull-down experiments and competition experiments using the native p85 biding site in the PDGFR, the dual-phosphorylated peptide (pYT) displayed a higher affinity for p85 than did pY. Thus, dual phosphorylation on both tyrosine and threonine at the predicted ERK phosphorylation site in Gab1 results in a higher affinity for p85 binding than does tyrosine phosphorylation alone. The antibody used in these experiments can recognize both p85␣ and ␤, and the association of these isoforms with Gab1 is not well distinguished using the gel separation techniques presented. Therefore the association of either p85␣, p85␤, or both may be up-regulated in this setting. Conversely, the dual phosphorylation of Gab1 could result in a decreased affinity of Gab1 for an independent PI3K regulatory isoform such as p55.
To test whether this proposed dual phosphorylation of Gab1 actually results in an alteration in PI3K downstream signaling, we examined the effects of inhibition of ERK activation on Akt phosphorylation. Akt is a PH domain containing serine/threonine kinase that is recruited to the membrane by the PI3K product PI 3,4,5 P 3 where it is phosphorylated and activated by the membrane-associated kinase PDK1, resulting in enhanced cell survival (28). Using an antibody specific for the activated state of Akt (anti-pT 308 ), we found that activated Akt was detectable within 2 min following HGF stimulation. In the absence of U0126 treatment, the phosphorylation state of Akt increased further at 5 and 10 min, consistent with a further increase in PI 3,4,5 P 3 production at the membrane. In the setting of ERK inhibition, the initial increase in phospho-Akt was not altered, but the subsequent further increase at 5 and 10 min was entirely abolished.
Taken together, these data are most consistent with the model that c-Met activation results in Gab1 recruitment to the receptor followed by phosphorylation of Gab1 at multiple tyrosine residues, including 472 YVPM 475 . This results in association of PI3K with Gab1 and local production of PI 3,4,5 P 3 . In concert, HGF induces ERK activation with the subsequent association of Gab1 and ERK resulting in phosphorylation of Gab1 at Thr 476 in the PI3K docking site. This dually phosphorylated motif then can act as a higher affinity site for the p85 subunit of the PI3K, leading to a further recruitment and/or stabilization of PI3K at the membrane and subsequently greater activation of PI 3,4,5 P 3 -dependent signaling pathways. The increase in Akt phosphorylation observed in the absence of ERK inhibition suggests yet another pathway whereby ERK activation may help promote cell survival and prevent apoptosis.
In conclusion, the HGF-mediated interaction of Gab1 and PI3K is dependent on tyrosine phosphorylation by the c-Met receptor, but is further enhanced by ERK-mediated phosphorylation of Gab1, possibly at Thr 476 . Threonine phosphorylation at this site results in a higher affinity of the PI3K binding site 472 YVPMTP 477 for p85, whereas inhibition of ERK activation decreased the p85 association with Gab1 and decreased the activation of the downstream kinase Akt. Because activation of the PI3K is critical for epithelial cell migration and morphogenesis, this result may in part explain our prior observations that HGF-mediated epithelial cell migration and tubule formation are dependent on MAPK activation (22).