Heregulin-dependent activation of phosphoinositide 3-kinase and Akt via the ErbB2/ErbB3 co-receptor.

The ErbB2/ErbB3 heregulin co-receptor has been shown to couple to phosphoinositide (PI) 3-kinase in a heregulin-dependent manner. The recruitment and activation of PI 3-kinase by this co-receptor is presumed to occur via its interaction with phosphorylated Tyr-Xaa-Xaa-Met (YXXM) motifs occurring in the ErbB3 C terminus. In this study, mutant ErbB3 receptor proteins expressed in COS7 cells were used to investigate PI 3-kinase-dependent signaling pathways activated by the ErbB2/ErbB3 co-receptor. We observed that a mutant ErbB3 protein with each of its six YXXM motifs containing a Tyr --> Phe substitution was unable to bind either the p85 regulatory or p110 catalytic subunit of PI 3-kinase. However, restoration of a single YXXM motif was sufficient to mediate association with the PI 3-kinase holoenzyme, although at a lower level than wild-type ErbB3. When ErbB3 YXXM motifs were restored in pairs, evidence for cooperativity between two, those incorporating Tyr-1273 and Tyr-1286, was observed. Interestingly, we have shown that an apparent association of PI 3-kinase activity with ErbB2/Neu was due to the residual presence of ErbB3 in ErbB2 immunoprecipitates. The necessity of ErbB3 association with PI 3-kinase for downstream signaling to the effector kinase Akt was also investigated. Here, the heregulin-dependent translocation of Akt to the plasma membrane and its subsequent activation was observed in intact NIH-3T3 fibroblasts. Recruitment of PI 3-kinase to ErbB3 was required for both activities, and it appeared that ErbB2 activation alone was not sufficient to activate PI 3-kinase signaling in these cells.

In particular, ErbB3 has been characterized as a major mediator of heregulin-dependent activation of the phosphoinositide (PI) 1 3-kinase pathway (6 -14). One general mechanism of recruitment and activation of PI 3-kinase involves the binding of the tandem Src homology 2 (SH2) domains of its p85 regulatory subunit to phosphorylated YXXM motifs found in signaling proteins (15,16). ErbB3 is particularly well adapted to mediate PI 3-kinase signaling, because it contains in its Cterminal phosphorylation domain six such consensus p85 binding motifs (17). Two previous studies have investigated the role of these YXXM motifs in ErbB3 signaling. One study showed that phosphopeptides containing the various ErbB3 YXXM motifs could inhibit p85 association with ErbB3 (9). Additionally, we have shown via the yeast two-hybrid system that these phosphorylated motifs can directly associate with the SH2 domains of p85 (18). However, it is not known whether ErbB3 YXXM motifs are solely responsible for the activation of PI 3-kinase by the ErbB2/ErbB3 co-receptor complex (18).
ErbB2/Neu has also been observed to associate with PI 3-kinase (8,11,19,20). Additionally, ErbB2 monoclonal antibodies inhibit heregulin-dependent activation of PI 3-kinase and its downstream target, Akt, in certain breast cancer cell lines (21). It is hypothesized that a YXXM motif occurring in the ErbB2 protein-tyrosine kinase domain could mediate a direct interaction with PI 3-kinase (11). Thus, the mechanism by which ErbB2 and ErbB3 activate PI 3-kinase is likely to be complex. Additionally, growth factor-mediated activation of PI 3-kinase has been observed to involve multiple events: p85 association with phosphorylated tyrosine residues, membrane localization, and Ras association with the p110 catalytic subunit (16,(22)(23)(24). The actions of other signaling proteins, such as RasGAP and phospholipase C␥, may negatively affect PI 3-kinase activity as in the case of platelet-derived growth factor (PDGF)-dependent signaling (24).
The insulin receptor substrate-1 (IRS-1), Drosophila insulin receptor, PDGF receptor, and hepatocyte growth factor receptor are additional examples of signaling proteins that, like ErbB3, contain multiple YXXM motifs. IRS-1 has been shown to become phosphorylated by the insulin receptor and subsequently associate with and activate PI 3-kinase (25,26). Several in vitro studies have examined the association of p85 with IRS-1-derived phosphotyrosyl peptides containing YXXM motifs. The high affinity association of p85 with IRS-1 YXXM phosphopeptides was observed to depend on the specific YXXM sequence order, because random shuffling of the tyrosine and methionine residues did not allow for phosphopeptide association (27,28). Additionally, the phosphorylated IRS-1 YXXM motifs regulated a conformational change in the SH2 domains of p85 that was associated with increased PI 3-kinase activity (29). Interestingly, in vitro phosphopeptide studies suggested cooperation between the multiple IRS-1 YXXM motifs. Bisphosphorylated YXXM peptides had a significantly increased affinity for tandem p85 SH2 domains as compared with monophosphorylated peptides (30). Also, IRS-1 bisphosphorylated peptides elicited a 2-fold higher activation of PI 3-kinase activity relative to monophosphorylated peptides (16).
The Drosophila insulin receptor protein contains three YXXM motifs, whereas the human IR contains only one (31). Therefore, the signaling capabilities of these receptors have been previously compared. However, from these comparative studies, it has not become clear as to how the additional YXXM motifs in Drosophila insulin receptor contribute to signaling by this receptor (32,33).
The SH2 domains of p85 have been shown to be required for the association of PI 3-kinase with PDGF receptor (15). Two tyrosines occurring in the PDGF receptor, Tyr-740 and Tyr-751, are in the context of YXXM motifs. To investigate the ability of these sites to associate with PI 3-kinase, a mutant PDGF receptor with phenylalanine substitutions at Tyr-740 and Tyr-751 was generated. It was observed that the PDGF receptor Y740F-Y751F mutant was unable to associate with PI 3-kinase (34). Interestingly, when the tyrosine residues were individually restored as single add-back mutants, the PDGF receptor Tyr-740 and Tyr-751 mutants were able to bind PI 3-kinase activity only 3% or 23%, respectively, of that bound by the wild-type receptor. These results suggested that both the Tyr-740 and Tyr-751 sites were required for PI 3-kinase to maximally associate with the PDGF receptor. Dual phosphorylation sites also appear to be critical for hepatocyte growth factor receptor interaction with PI 3-kinase (35).
The PI 3-kinase-dependent generation of PtdIns 3,4-bisphosphate and PtdIns 3,4,5-trisphosphate has been shown in several systems to result in the translocation of the Ser/Thr kinase Akt to the plasma membrane, where it is activated via its phosphorylation by PtdIns phosphate-dependent protein kinases (PDKs) (36 -38). Akt is recruited to the plasma membrane by the binding of its pleckstrin homology domain (PH Akt ) to PtdIns 3,4-bisphosphate or PtdIns 3,4,5-trisphosphate (39 -45), which leads to the phosphorylation of Akt on Thr-308 and Ser-473 by PDK1 and PDK2 (46 -49). Phosphorylation on Thr-308 and Ser-473 is critical for the activation of the protein Ser/Thr kinase activity of Akt. Interestingly, PDK1 and PDK2 also bind PtdIns 3,4,5-trisphosphate and are postulated to translocate to the plasma membrane (37). Because of its high degree of regulation by D-3-phosphorylated PtdIns species, the activation of Akt provides an excellent indication of the production of PI 3-kinase products in the cell. Indeed, by fusion of the green fluorescent protein (GFP) module to PH Akt , the translocation of PH Akt to the plasma membrane can be observed in live cells by fluorescence microscopy (50,51).
Given this background, we posed several questions about the mechanism of heregulin-dependent PI 3-kinase activation via ErbB2 and ErbB3 YXXM motifs. First, is a single YXXM motif necessary and sufficient to mediate PI 3-kinase association with the ErbB2/ErbB3 co-receptor? Second, which of these various YXXM motifs can interact with PI 3-kinase? Third, can two or more of the six ErbB3 YXXM motifs cooperate in PI 3-kinase binding and activation as in the case of the PDGF receptor? We also sought to determine whether a proline-rich sequence in the C terminus of ErbB3 that potentially binds the SH3 domain of p85 (18) participates in PI 3-kinase activation.
To address these questions about the PI 3-kinase activation mechanism, ErbB2 and ErbB3 receptors with specific amino acid substitutions were generated by site-directed mutagenesis and transiently expressed in COS7 cells. NIH-3T3 cells lines stably expressing wild-type and mutant receptors were also employed in investigations of downstream signaling by PI 3-kinase. Here, we examined the heregulin-dependent activation of the PI 3-kinase target, Akt, and the production of PI 3-kinase products in intact cells.
Plasmids-The rat ErbB3 expression plasmid, pcDNA3-B3, has been described previously (17). The rat ErbB2/Neu expression plasmid, pcDNA3-B2, was constructed by cloning the rat ErbB2 cDNA, a generous gift of Dr. Robert Weinberg (Whitehead Institute, Cambridge, MA), into pcDNA3 (Invitrogen). Mutant receptor cDNAs were generated by Ex-Site polymerase chain reaction mutagenesis (Stratagene). A codon corresponding to Tyr-952 in the ErbB2 cDNA sequence was mutated to phenylalanine to generate the ErbB2-952F cDNA by using pcDNA3-B2 as a template. Seven codons corresponding to the loss of amino acids 1205-1211 (PPRPPRP) were deleted from pcDNA3-B3 to generate the ErbB3⌬Pro cDNA. The ErbB3-6F cDNA was constructed with pcDNA3-B3 as a template by six sequential rounds of polymerase chain reaction mutagenesis by which each codon corresponding to Tyr-1051, Tyr-1194, Tyr-1219, Tyr-1257, Tyr-1273, and Tyr-1286 was substituted with a phenylalanine codon. ErbB3-1051Y, ErbB3-1194Y, ErbB3-1219Y, ErbB3-1257Y, ErbB3-1273Y, or ErbB3-1286Y add-back mutant cDNAs were generated with the ErbB3-6F cDNA as a template. Each of the sequences altered by Ex-Site mutagenesis were subcloned into the parent expression vector and sequenced to verify that the desired mutation was incorporated without the introduction of additional sequence alterations.
Cell Culture and Electroporation-COS7 cells were purchased from and maintained as recommended by the American Type Culture Collection. COS7 cells were transiently transfected using specific combinations of expression plasmids as indicated by electroporation at 210 volts in serum-free DMEM with a Cell-Porator (Life Technologies, Inc.). Electroporated cells were subsequently plated on 100-mm dishes in DMEM supplemented with 10% fetal bovine serum and 0.01% gentamicin. Twenty-four hours after plating, the cells were washed in DMEM and serum-starved in DMEM supplemented with 0.1% fetal bovine serum for an additional 24 h before heregulin stimulation.
PI 3-Kinase Assays-The lysates were incubated with ErbB2 or ErbB3 antibody as indicated for 1 h at 4°C. Each antibody-lysate mixture was added to 200 l of protein G-agarose (1:6 suspension) and rocked for 1 h. The protein G-agarose slurry was washed three times in Nonidet P-40 lysis buffer and resuspended in 1 ml of lysis buffer. 0.5 ml of the agarose slurry was used for Western blotting, and the remaining 0.5 ml was used for PI 3-kinase analysis. For PI 3-kinase assays (52), the immunoprecipitate was additionally washed two times with buffer 2 (100 mM Tris/HCl, 5 mM LiCl, 100 M sodium vanadate, 100 mM sodium fluoride, 10 mM sodium pryrophosphate, and 2 mM dichloroace-tic acid, pH 7.4), washed two times with buffer 3 (10 mM Tris/HCl, 150 mM NaCl, 5 mM EDTA, 100 M sodium vanadate, 100 mM sodium fluoride, 10 mM sodium pryrophosphate, and 2 mM dichloroacetic acid, pH 7.4), and resuspended in 65 l of 10 mM Tris/HCl (pH 7.4) containing 150 mM NaCl, 5 mM EDTA, 4 mM MgCl 2 , 2.3 mg/ml phosphotidylserine, and 1.5 mg/ml phosphotidylinositol. The reaction was initiated by adding 10 l of 100 M ATP containing 10 Ci of [␥-32 P]ATP (3000 Ci/ mmol), 10 mM MgCl 2 , 200 M adenine, and 45 mM HEPES/Na (pH 7.8). The reaction was stopped by adding 20 l of 8 N hydrochloric acid. Lipids were extracted with 150 l of methanol/chloroform (1:1), and 50 l of the extract was resolved on potassium oxalate-coated silica thin layer chromatography plates with chloroform/methanol/water/ammonium hydroxide (60:47:11:2) as the solvent. The phosphatidylinositol 3-phosphate products were detected by autoradiography, scraped from plates, and quantified by liquid scintillation counting or detected and quantified with a PhosphorImager (Molecular Dynamics).
Western Immunoblotting-Cell lysate samples and immunoprecipitates were resuspended in gel sample buffer, resolved by SDS-polyacrylamide gel electrophoresis, and transferred to Immobilon membranes (Millipore). Membranes were blocked for 1 h in 5% (w/v) dry milk, PBS, and 0.1% Tween 20. Primary antibodies as indicated were incubated with membranes for 1 h, and the membranes were washed in PBS with 0.1% Tween 20. Subsequently, horseradish peroxidase-conjugated secondary antibodies were added for 1 h, and the membranes were washed in PBS with 0.1% Tween 20. Super Signal (Pierce) enhanced chemiluminescence reagent was used to detect membrane-bound protein by luminography.
In Vivo Fluorescence Microscopy Assay of PI 3-Kinase Activation-An RNA expression vector encoding a GFP-Akt-PH domain fusion protein (GFP-PH Akt ) has previously been described by Kontos et al. (50). The cDNA encoding GFP-PH Akt was subcloned from this vector into the mammalian expression vector pcDNA3. NIH-3T3 cells lines expressing ErbB3 receptors were plated in chamber slides (Lab-Tek) and transfected with pcDNA3-GFP-PH Akt via the LipofectAMINE transfection method. Twenty-four hours after transfection, GFP-PH Akt distribution in cells was visualized at 100ϫ magnification by confocal laser scanning fluorescence microscopy (Zeiss 510). The cells were then stimulated with 5 nM heregulin for 10 min, and the original cell field was again visualized.

Role of ErbB3 YXXM and Proline-rich Consensus Binding Motifs in PI 3-Kinase Association-ErbB3
YXXM motifs have previously been shown by phosphopeptide competition assays (9) and yeast two-hybrid system analysis (18) to be critical for ErbB3 association with p85. In this study, the association of p85 and PI 3-kinase activity with ErbB3 YXXM motifs in the context of intact cultured cells was examined by site-directed mutagenesis. Six consensus binding motifs for p85 occurring in the C terminus of ErbB3 were mutated from Tyr-Xaa-Xaa-Met (YXXM) to Phe-Xaa-Xaa-Met (FXXM) by mutagenesis of the corresponding cDNA sequence to generate the ErbB3-6F expression plasmid. Previously, we had detected an in vitro interaction between ErbB3 and the SH3 domain of p85, which we suspected to be dependent on a proline-rich consensus binding motif (PPRPPRP) occurring in the ErbB3 C terminus (18). Therefore, we also deleted the ErbB3 cDNA sequence encoding the amino acid sequence PPRPPRP to generate an ErbB3⌬Pro expression plasmid.
We transiently expressed ErbB3-WT, ErbB3-6F, and ErbB3⌬Pro together with ErbB2 in COS7 cells and examined the ability of these receptors to associate with PI 3-kinase upon stimulation by heregulin. Immunoprecipitation analysis revealed that ErbB3-WT showed heregulin-dependent association with p85 (Fig. 1B). In contrast, ErbB3-6F associated negligibly with p85 as compared with the wild-type receptor (Fig.  1B), although expression of both receptors was comparable (Fig. 1A). Examination of the relative phosphorylation levels of FIG. 1. PI 3-kinase association with wild-type and mutant ErbB3 proteins. COS7 cells were transiently cotransfected with the pcDNA3 vector incorporating the ErbB2 cDNA and either the parent expression vector (Ϫ) or the vector incorporating wild-type ErbB3 (WT), ErbB3⌬Pro (⌬Pro), or ErbB3-6F (6F) cDNA as indicated. Transfected cells were treated for 20 min in the absence (Ϫ) or presence (ϩ) of 1 nM heregulin (HRG) and subjected to detergent lysis. A, cell lysates were immunoblotted (IB) with antibodies recognizing ErbB2 (␣-B2), ErbB3 (␣-B3), and p85 (␣-p85) as indicated. B, lysates were immunoprecipitated with ErbB3-specific antibody. Samples of each immunoprecipitate were immunoblotted with antibodies recognizing ErbB3, phosphotyrosine (␣-P-tyr), and p85. C, one-half of each ErbB3 immunoprecipitate was analyzed for PI 3-kinase activity by a thin layer chromatographic assay. D, ErbB3 immunoprecipitates described in B were immunoblotted with antibodies recognizing the three Shc isoforms. These results are representative of three independent experiments. the receptors showed a heregulin-dependent enhancement in phosphorylation of both ErbB3-WT and ErbB3-6F (Fig. 1B). Thus, it appeared that the wild-type and mutant receptors were functionally expressed at the cell surface and that some of the six remaining tyrosine residues occurring in the C terminus of ErbB3-6F were subject to phosphorylation. PI 3-kinase activity in ErbB3 immunoprecipitates was analyzed by in vitro kinase assays in which the PtdIns 3-phosphate product was resolved by thin layer chromatography. In accordance with p85 immunoblotting analysis, PI 3-kinase activity did not appreciably co-immunoprecipitate with ErbB3-6F when compared with the wild-type receptor (Fig. 1C). Therefore, it appeared that ErbB3 tyrosine residues occurring in YXXM motifs were principally responsible for the PI 3-kinase interaction with ErbB3, presumably via p85 SH2 domains.
ErbB3⌬Pro immunoprecipitates were also analyzed for p85 binding and PI 3-kinase activity to examine the potential role of the PPRPPRP motif in PI 3-kinase signaling. ErbB3⌬Pro was unimpaired, as compared with ErbB3-WT, in its abilities to bind p85 and to immunoprecipitate PI 3-kinase activity (Fig.  1, B and C). Thus, the proline-rich consensus binding site for the SH3 domain of p85 was not necessary for the recruitment of PI 3-kinase by ErbB3.
ErbB3-WT, ErbB3-6F, and ErbB3-⌬pro all associated with Shc adapter proteins (Fig. 1D), which was consistent with these receptors retaining the NPDY consensus binding site for Shc. Additional studies in our laboratory have shown that ErbB3-6F also retains the ability to activate the Ras-mitogen-activated protein kinase pathway. 2 Comparison of PI 3-Kinase Activity in ErbB2 and ErbB3 Immunoprecipitates-It has been observed in some studies (8,11,19), but not others (10,14), that p85 co-immunoprecipitates from cell lysates with ErbB2. Here, we transiently transfected COS7 cells with ErbB2 and ErbB3 and compared the abilities of ErbB2 and ErbB3 antibodies to co-immunoprecipitate p85 and PI 3-kinase activity following stimulation by heregulin. Both p85 and PI 3-kinase activity were detected in ErbB2 and ErbB3 immunoprecipitates (Fig. 2). However, the ability of the ErbB3 antibody to co-immunoprecipitate p85 and PI 3-kinase activity was considerably greater than that of the ErbB2 antibody. Although we have detected negligible amounts of ErbB3 in ErbB2 immunoprecipitates (data not shown), we considered the possibility that residual amounts of ErbB3 co-immunoprecipitated with ErbB2, which could have been responsible for the low level of PI 3-kinase activity immunoprecipitated with the ErbB2 antibody.
Comparison of PI 3-Kinase Activity in ErbB2 Immunoprecipitates from Cells Containing ErbB2 and ErbB3 YXXM 3 FXXM Mutant Receptors-It has been suggested that a single YXXM motif (Tyr-952) found within the protein-tyrosine kinase domain of ErbB2 is capable of being phosphorylated (53) and could subsequently mediate p85 association with ErbB2. To test this possibility, we mutated ErbB2 Tyr-952 to phenylalanine and co-expressed the mutant receptor, ErbB2-952F, with ErbB3. We observed that the Tyr-952 3 Phe mutation had no effect on the immunoprecipitation of PI 3-kinase activity with ErbB2 (Fig. 3A). We also expressed ErbB3-6F with either ErbB2 or ErbB2-952F. In both cases, we observed that ErbB2 and ErbB2-952F were unable to co-immunoprecipitate PI 3-kinase activity (Fig. 3A). In all cases, comparable amounts of phosphorylated ErbB2 were immunoprecipitated (Fig. 3B). Thus, it appeared that the relatively low levels of p85 and PI 3-kinase activity that immunoprecipitated with ErbB2 were due to the residual co-immunoprecipitation of ErbB3 with the ErbB2 co-receptor.

Comparison of ErbB3
Single YXXM Add-back Mutants-Our initial goal was to determine which of the six consensus YXXM motifs occurring in the ErbB3 C terminus were capable of interacting with PI 3-kinase. We generated Phe 3 Tyr addback mutants using the ErbB3-6F protein as a template to create a panel of ErbB3 receptors containing a restored YXXM motif at amino acid residue 1051, 1194, 1219, 1257, 1273, or 1286. ErbB3-WT, ErbB3-6F, and each of the ErbB3 YXXM add-back mutants (ErbB3-1051Y, ErbB3-1194Y, ErbB3-1219Y, ErbB3-1257Y, ErbB3-1273Y, and ErbB3-1286Y) were transiently expressed with ErbB2 in COS7 cells. The transfected cells were stimulated with heregulin and lysed, and the ErbB3 immunoprecipitates were analyzed. Comparable expression and phosphorylation of the various receptors were observed (Fig. 4A). The ability of the ErbB3 receptors to associate with the two subunits of PI 3-kinase, the p85 regulatory subunit and the p110 catalytic subunit, was examined by immunoblotting ErbB3 immunoprecipitates with p85 and p110 antibodies, respectively. We observed that ErbB3-6F, which was devoid of YXXM motifs, was unable to associate with either p85 or p110 (Fig. 4B). Unlike ErbB3-6F, the various ErbB3 add-back mutants each associated with the two subunits of PI 3-kinase (Fig.  4B). Although each mutant receptor bound less p85 and p110 than did ErbB3-WT, there appeared to be only minor differences among the mutants in their ability to co-immunoprecipi- tate p85 and p110 (Fig. 4B). However, in vitro PI 3-kinase activity analysis revealed significant differences between the mutants in their abilities to activate PI 3-kinase. ErbB3-WT, ErbB3-1051Y, and ErbB3-1194Y bound substantially more PI 3-kinase activity relative to ErbB3-6F (Fig. 4C). Whereas ErbB3-1051Y and ErbB3-1194Y showed associated PI 3-kinase activities one-half and one-third respectively, of that of the wild-type receptor, ErbB3-1219Y, ErbB3-1257Y, ErbB3-1273Y, and ErbB3-1286Y exhibited very low to negligible associations with PI 3-kinase activity, which were comparable with that of ErbB3-6F (Fig. 4C). These results suggested that PI 3-kinase was activated differentially by the various YXXM sites occurring in ErbB3. Surprisingly, PI 3-kinase activity did not correlate with the ability of each of these sites to associate with the p85/p110 subunit complex.
Comparison of Single and Double Add-back Mutants-There is evidence that phosphopeptides containing two phosphorylated tyrosine residues have a higher binding affinity for p85 than monophosphorylated peptides, which is presumably due to their interaction with the tandem SH2 domains of p85 (see the Introduction). Given the multiple YXXM motifs in ErbB3, we examined the possibility that pairs of ErbB3 YXXM motifs cooperate in binding p85. To this end, we generated ErbB3 double add-back mutants that contained pairs of restored YXXM motifs: ErbB3-1051/1286Y, ErbB3-1051/1194Y, and ErbB3-1273/1286Y. The corresponding single add-back mutants and ErbB3-6F were used for comparison.
We first determined levels of receptor expression and phosphorylation for each pair of YXXM motifs investigated (Tyr-1051/1286, Tyr-1051/1194, and Tyr-1273/1286). Representative results are shown for the single add-back mutants ErbB3-1051Y and ErbB3-1286Y and the ErbB3-1051/1286Y double add-back mutant (Fig. 5A). Each of the receptors was expressed to a similar degree as determined by immunoblotting, although there were slight differences in the levels of receptor phosphorylation. Receptor-associated p85 was also assayed for each set of sites, as represented in Fig. 5B for ErbB3-1051/1286Y. In general, double add-back mutants showed increased heregulin-dependent p85 association in comparison with single add-back mutants (Fig. 5B). As shown previously, p85 association with ErbB3-6F was negligible.
Receptor-associated PI 3-kinase activity was determined by in vitro kinase assays as described under "Experimental Procedures." Multiple experiments revealed that ErbB3-1051Y

FIG. 3. PI 3-kinase catalytic activity in ErbB2 immunoprecipitates from cells expressing wild-type or mutant ErbB receptors.
COS7 cells co-transfected with the indicated ErbB2 or ErbB3 expression plasmids were treated for 20 min with 1 nM heregulin. The cells were lysed, and the lysates were immunoprecipitated with ErbB2 antibody. A, aliquots of the immunoprecipitates were analyzed for PI 3-kinase activity by a thin layer chromatographic assay. B, aliquots of ErbB2 immunoprecipitates (IP) were immunoblotted (IB) with antibodies recognizing phosphotyrosine. The results shown are representative of three independent experiments.

FIG. 4. PI 3-kinase association with wild-type and single YXXM add-back mutant ErbB3 proteins.
COS7 cells were transiently cotransfected with the pcDNA3 vector incorporating the ErbB2 cDNA and with the vector incorporating ErbB3-WT, ErbB3-1051Y, ErbB3-1194Y, ErbB3-1219Y, ErbB3-1257Y, ErbB3-1273Y, ErbB3-1286Y, or ErbB3-6F cDNA as indicated. The transfected cells were treated for 20 min with 1 nM heregulin and subjected to detergent lysis. The cell lysates were then immunoprecipitated with ErbB3-specific antibody (␣-B3). A, immunoprecipitates (IP) were immunoblotted (IB) with antibodies recognizing ErbB3 and phosphotyrosine (␣-p-Tyr) as indicated. B, one-half of each immunoprecipitate was immunoblotted with antibodies recognizing p85 (␣-85) or p110 (␣-p110). Expression levels of p85 in lysates were examined by immunoblotting with ␣-p85. C, the remainder of each ErbB3 immunoprecipitate was analyzed for PI 3-kinase activity by a thin layer chromatographic assay. Shown graphically are the results of three independent experiments normalized to the PI 3-kinase activity associated with ErbB3-WT. The bars shown indicate the standard error. and ErbB3-1051/1286Y were roughly equivalent in their ability to bind PI 3-kinase activity, despite the increased p85 association observed with ErbB3-1051/1286Y (Fig. 5C). ErbB3-1286Y bound negligible enzyme activity (Fig. 5C), despite its ability to bind p85 (Fig. 5B). These results were consistent with those of Fig. 4C, where ErbB3-1051Y was seen to mediate a much greater heregulin-dependent increase in PI 3-kinase activity than was observed with ErbB3-1286Y. Receptor-associated p85 protein and PI 3-kinase activity were also evaluated for the ErbB3-1051/1194Y and ErbB3-1273/1286Y double add-back receptors and their single add-back counterparts, with the analysis of these results described below.
To determine whether the interaction of receptor-associated PI 3-kinase with paired YXXM motifs could enhanced the intrinsic activity of the enzyme, activities from in vitro kinase assays (Fig. 5C) were first normalized relative to the amount of p85 bound (Fig. 5B). We then compared the specific activities of PI 3-kinase associated with pairs of single add-back mutant receptors with that associated with the corresponding double add-back mutants (Fig. 5D). When PI 3-kinase activities were normalized to the amount of p85 bound to the receptors, ErbB3-1051/1286Y was less efficient in activating PI 3-kinase than was ErbB3-1051Y (Fig. 5D). This suggested that PI 3-ki-nase activation was due solely to its association with ErbB3 Tyr-1051. No evidence for cooperativity between the two sites in ErbB3-1051/1286Y was observed.
Additional comparisons between paired sites revealed that ErbB3-1051Y and ErbB3-1194Y were roughly equivalent in their abilities to activate PI 3-kinase activity, with ErbB3-1051/1194Y showing an intermediate activation (Fig. 5D). There was again no evidence for cooperativity between sites. In contrast, normalized receptor-associated PI 3-kinase activities for both ErbB3-1273Y and ErbB3-1286Y were low (Fig. 5D). Although these sites individually demonstrated weak activation, they are closely spaced and could potentially cooperate in activating the enzyme. Indeed, the intrinsic activity of PI 3-kinase bound to the double add-back mutant ErbB3-1273/1286Y was greater than that bound to either single add-back receptor mutant (Fig. 5D), which indicated that the two ErbB3 YXXM motifs cooperated in activating the bound enzyme. From these three paired site comparisons, it appeared that closely spaced tyrosine residues, such as Tyr-1273 and Tyr-1286, could cooperate in activating PI 3-kinase, whereas more widely separated pairs of sites (e.g. Tyr-1051/1194 and Tyr-1051/1286) functioned independently (see "Discussion").

FIG. 5. PI 3-kinase association with single and double YXXM add-back mutant proteins. COS7 cells were transiently co-transfected
with the pcDNA3 vector incorporating the ErbB2 cDNA and vector incorporating ErbB3-1051Y, ErbB3-1286Y, ErbB3-1051/1286Y, or ErbB3-6F cDNA as indicated. Transfected cells were stimulated with heregulin and lysed, and the ErbB3 protein was immunoprecipitated (see Fig. 4). A, ErbB3 immunoprecipitates (IP) were immunoblotted (IB) with antibodies recognizing ErbB3 (␣-B3) and phosphotyrosine (␣-P-Tyr) as indicated. B, immunoprecipitate and lysate samples were immunoblotted with antibody recognizing p85 (␣-p85). C, the remaining half of each ErbB3 immunoprecipitate was analyzed for PI 3-kinase activity by a thin layer chromatographic assay. D, as described above, COS7 cells transiently expressing ErbB2 and one of the various single and double add-back ErbB3 mutants or ErbB3-6F were stimulated with heregulin and analyzed for receptor expression, receptor phosphorylation, and receptor-associated p85 protein and PI 3-kinase activity. PI 3-kinase activities were normalized to the level of ErbB3-associated p85 protein as determined by densitometric analysis of ␣-p85 immunoblots, as shown in B. PI 3-kinase specific activities for single add-back mutants were expressed relative to those of the corresponding double add1-back mutants, and the results of four independent experiments were averaged. Open bars indicate N-terminal single YXXM add-backs (ErbB3-1051Y or ErbB3-1273Y as designated), hatched bars indicate C-terminal single YXXM add-backs (ErbB3-1286Y or ErbB3-1194Y), and solid bars indicate double YXXM add-backs (ErbB3-1051/1286Y, ErbB3-1051/1194Y, or ErbB3-1273/1286Y). The bars indicate the standard error.

ErbB3-mediated PI 3-Kinase Signaling in Intact Cells-
The experiments described above explored the mechanism of recruitment of PI 3-kinase by the activated ErbB3 receptor. Presumably, in the cellular context, the observed phosphotyrosinedependent association of PI 3-kinase with ErbB3 would lead to an elevation of the levels of PtdIns 3,4-bisphosphate and Pt-dIns 3,4,5-trisphosphate, which have well characterized second messenger functions. The aim of our next experiments was to assay the ErbB3-mediated generation of such second messengers in the intact cell.
To characterize the capacity of the ErbB3 receptor to activate PI 3-kinase in intact cells, NIH-3T3 fibroblast cell lines expressing endogenous ErbB2 and either the ErbB3-WT or ErbB3-6F receptor protein were transiently transfected with a vector expressing a GFP-PH Akt fusion protein. Thereafter, the subcellular localization of expressed GFP-PH Akt was monitored prior to and after stimulation of the cells with heregulin (Fig.  6). In cells expressing the wild-type ErbB3 protein, heregulin clearly stimulated the translocation of GFP-PH Akt from the cytoplasm to the plasma membrane. No heregulin-dependent GFP-PH Akt translocation was seen in vector-transfected cells or ErbB3-6F-expressing cells (Fig. 6). These results demonstrated the heregulin-and ErbB3-dependent activation of PI 3-kinase in intact cells and were consistent with the above demonstration that ErbB3 YXXM motifs are necessary for PI 3-kinase recruitment to the ErbB2/ErbB3 co-receptor. Notably, it appeared that endogenous ErbB2, albeit expressed in NIH-3T3 cells at only a moderate level, 2 was not itself capable of mediating significant PI 3-kinase activation when expressed with ErbB3-6F.
Analysis of Akt Activation via ErbB2/ErbB3 Co-receptors-In addition to Akt translocation, we also investigated Akt phosphorylation on Ser-473, which presumably regulates the catalytic activity of Akt. Cell lines expressing the ErbB3-WT or ErbB3-6F receptor were stimulated with heregulin, and Akt phosphorylation was detected via a phospho-specific Akt anti-body. We observed a heregulin-dependent increase in phosphorylation of Akt on Ser-473 in cells expressing ErbB3-WT receptor, but not vector-control or ErbB3-6F-expressing cell lines (Fig. 7A). These results corroborated our GFP-PH Akt translocation assays in intact cell lines.
We next investigated whether the translocation and phosphorylation of Akt on Ser-473 would be associated with an increase in the catalytic activity of Akt. To investigate Akt activation, cell lines expressing the ErbB3-WT or ErbB3-6F receptor were stimulated with heregulin, Akt was immunoprecipitated, and in vitro kinase assays were performed using GSK3␣ as a substrate. We observed that heregulin stimulated an increase in Akt activity in cells expressing ErbB3-WT (Fig.  7B). However, neither ErbB3-6F-expressing nor vector control cells showed a heregulin-dependent increase in Akt activity, which was presumably due to the inability of ErbB3-6F to activate PI 3-kinase. From these experiments, it can be concluded that the recruitment of PI 3-kinase by ErbB3 is necessary not only for heregulin-dependent Akt translocation but also for Akt kinase activation. DISCUSSION In this study, the role of six ErbB3 YXXM motifs in PI 3-kinase signaling was investigated. To this end, we generated an ErbB3 mutant receptor (ErbB3-6F) devoid of the ability to associate with the p85 subunit of PI 3-kinase by mutation of six tyrosine residues occurring in YXXM consensus binding sites for p85. Our observation that YXXM consensus motifs were required for the in vivo association of p85 with ErbB3 agreed with previous in vitro (9) and yeast two-hybrid system (18) analyses.
Additionally, the role of an ErbB3 proline-rich binding motif in PI 3-kinase signaling was investigated. It seemed plausible that the ErbB3 proline-rich consensus binding sequence for the SH3 domain of p85 might mediate an ErbB3/p85 interaction, given its sequence similarity to high affinity peptide ligands for this domain (54). Indeed, we had previously characterized the ErbB2/ErbB3 Signaling via PI 3-Kinase in vitro association of the p85 SH3 domain and a recombinant ErbB3 protein (18). However, we have shown here via an ErbB3 deletion mutant, ErbB3⌬Pro, that the potential p85 SH3 domain interaction with this proline-rich sequence was not necessary for PI 3-kinase association. It was observed that ErbB3-6F did not appreciably interact with p85, which suggested that the proline-rich region was also not sufficient to mediate high affinity binding. It is possible that the prolinerich motif of ErbB3 is not accessible to the SH3 domain of p85 in the cell. Alternatively, an intramolecular interaction of this SH3 domain with a known proline-rich consensus motif in p85 (55) could compete with any ErbB3/SH3 domain interaction and negate its significance.
The ErbB3-6F protein also proved useful in examining the potential association of the ErbB2 protein with PI 3-kinase. The ability of ErbB2 to associate with p85 and PI 3-kinase has been described (8,11,19). However, in murine 32D cells cotransfected with ErbB2 and ErbB3 cDNAs, ErbB3, but not ErbB2, was detected in p85 immunoprecipitates (10). Also, in transfected 293 and mammary carcinoma-derived MCF-7 cells, a heregulin-dependent increase in PI 3-kinase activity was attributed to ErbB3 but not ErbB2 (14). In our study, we observed the co-immunoprecipitation of PI 3-kinase activity with the ErbB2 protein, although it was much weaker than the co-precipitation seen with ErbB3. However, we concluded that the PI 3-kinase activity in ErbB2 immunoprecipitates was attributable to ErbB3 co-precipitating with ErbB2 as demonstrated with mutant ErbB receptors. Hence, the association of PI 3-kinase with ErbB2 and ErbB3 was observed only in the context of a wild-type ErbB3 receptor containing consensus p85 binding motifs. Our results perhaps explain some of the inconsistency in the literature regarding the ability of ErbB2 to associate with PI 3-kinase, although the possibility certainly exists for variability between experimental techniques and cell types used. Alternatively, ErbB3-6F could disrupt p85 interaction with ErbB2 or alter the pattern of ErbB2 phosphorylation.
ErbB2 has also been reported to associate with Src (53, 56), another potential mediator of PI 3-kinase activation (57), which presents another possible means by which ErbB2/ErbB3 coreceptors might signal through the PI 3-kinase pathway. However, in our study, PI 3-kinase signaling was not observed in vivo in the context of ErbB2/ErbB3-6F co-receptors (Figs. 6 and 7).
To further investigate the mechanism of p85 interaction with the ErbB3 protein, we created single and double YXXM motif add-back mutants of ErbB3. These mutant receptors allowed us to examine p85 binding to the six individual consensus binding sites occurring in ErbB3. We observed that each individual site was sufficient to mediate an interaction with PI 3-kinase and that at least one was required. In each case, the mutant receptors bound less p85 and associated with a lower level of PI 3-kinase activity than did the wild-type receptor. These studies suggested that each ErbB3 YXXM motif could be phosphorylated and associate with p85, although in the context of the wild-type receptor, the possibility of a preferential phosphorylation site usage certainly exists.
In studies of the single-site add-back mutants, we also examined whether the catalytic subunit of PI 3-kinase associated with each of the receptors. The p110 catalytic and p85 regulatory subunits were bound to a similar degree across the panel of add-back mutants but bound negligibly to ErbB3-6F. These results were consistent with the dimeric structure of the PI 3-kinase molecule. Surprisingly, the PI 3-kinase activity that was detected in ErbB3 immunoprecipitates did not correspond to the relative levels of p85 or p110 that were immunoprecipitated. There are at least three possibilities for the presence of PI 3-kinase with evidently low intrinsic activity in ErbB3 im-munoprecipitates: specificity of PI 3-kinase activation by the individual phosphopeptide recognition sequences, steric hindrances by the ErbB3 protein structure, or regulatory phosphorylation of PI 3-kinase.
In the case of the first possibility, it has been observed that both activating and nonactivating peptides can bind to the SH2 domain of p85 (29). Only the activating peptides can induce a distinct conformational change in the SH2 domain and increase p110 enzymatic activity (29). The various ErbB3 motifs could be split into categories of activating and nonactivating sequences. However, this would seem unlikely given that each sequence contains the canonical YXXM motif and appeared to have similar affinities as determined by immunoblotting analysis. Based on these observations, we would predict the various sequences to induce similar changes in PI 3-kinase activity (29). A second possibility is that steric hindrance presented by a given single add-back receptor mutant could inhibit conformational changes necessary for the activation of PI 3-kinase (29). This would be consistent with our observation that closely spaced sites activated enzyme activity more efficiently in tandem than individually (Fig. 5D).
A third possibility for the different levels of PI 3-kinase activity that associated with ErbB3 add-back mutants could be the autoregulation of the enzyme complex. Phosphorylation of Ser-608 in p85 by p110 has been observed to decrease the lipid kinase activity of the catalytic p110 subunit (58). Phosphatase 2A can reverse this inhibition, and it is hypothesized that phosphatases can disinhibit PI 3-kinase in a regulatory manner (58,59). Hypothetically, certain PI 3-kinase binding sites in ErbB3 could facilitate phosphatase access to p85. However, the phosphorylation state of Ser-608 in p85 was not determined in our study, and the mechanism of PI 3-kinase regulation by phosphorylation has not been well defined elsewhere in the literature.
We also investigated whether multiple ErbB3 YXXM motifs cooperated in binding and activation of PI 3-kinase by examining double add-back mutants in comparison to single add-back mutants. Although we observed increased p85 binding with the double add-back mutants compared with the respective single add-back mutants, this increased binding was in general not more than additive. However, if YXXM motifs did not cooperate in binding PI 3-kinase, they might yet have cooperated in activating the intrinsic catalytic activity of the bound enzyme. Hence, we compared the activities of bound PI 3-kinase after first normalizing these activities by the amount of bound p85 (Fig.  5D). In this analysis, the specific activity of PI 3-kinase bound to a double add-back mutant would, in the absence of cooperative activating effects, be intermediate to those of the single add-back mutants. Such was the case for the more widely spaced pairs of YXXM motifs examined (i.e. those in ErbB3-1051/1194Y and ErbB3-1051/1286Y). However, in the case of the C-terminal tandem pair, the double add-back mutant ErbB3-1273/1286Y showed activity somewhat higher than either single add-back mutant. Therefore, ErbB3 contains tandem sites (those incorporating Tyr-1273 and Tyr-1286) that appear to cooperate in a fashion analogous to the PDGF and hepatocyte growth factor receptors in the activation of PI 3-kinase but also contains individual sites (Tyr-1051 and Tyr-1194) that are alone capable of mediating strong PI 3-kinase activation.
The ErbB3-WT protein bound greater levels of p85 and p110 than did the add-back mutants. The wild-type receptor also associated with more PI 3-kinase activity. We speculate that this was due to an increased number of phosphorylation motifs and a greater opportunity for p85 binding sites to become phosphorylated. However, little is known regarding the in vivo stoichiometry of phosphorylation of the ErbB3 receptor. It appeared that each of the individual sites could become phospho-rylated, because each site could interact with p85. However, in the context of the wild-type receptor, some sites may be preferred over others. Alternatively, phosphorylation of ErbB3 by ErbB2 or other cytosolic kinases could be random. The stoichiometry of phosphorylation of the various sites was not determined here. Given our lack of knowledge of the stoichiometry of phosphorylation of individual YXXM motifs, it is difficult to make quantitative conclusions regarding cooperativity between YXXM motifs. However, given that it is likely that the stoichiometry of phosphorylation of the various double addback mutants employed in this study was significantly less than two phosphates/receptor molecule, it is probable that our analysis (Fig. 5D) underestimates the ability of ErbB3 PI 3-kinase binding motifs to act cooperatively in the binding and activation of this signaling enzyme.
Determination of ErbB3 phosphorylation patterns seems crucial to further address this problem. Nonetheless, we have shown here that each of the YXXM motifs was capable of being phosphorylated in the context of the single add-back receptors. Furthermore, it appeared that in the case of the wild-type receptor, there is an advantage in having a greater number of potential phosphorylation sites, because we observed an increased binding of PI 3-kinase to ErbB3-WT as compared with mutant receptors containing fewer available sites. Finally, we showed that two tandem phosphorylation sites in the extreme ErbB3 C terminus, although not efficiently activating PI 3-kinase independently, could cooperate in the activation of receptor-associated PI 3-kinase.
Our experiments with the GFP-PH Akt reporter in intact NIH-3T3 cells, along with Akt kinase assays, demonstrated that the six YXXM consensus binding sites in ErbB3 are necessary for the in vivo activation of PI 3-kinase signaling via the ErbB2/ErbB3 co-receptor. Recent studies in our laboratory have shown a heregulin-dependent activation of the Ras-mitogen activated protein kinase cascade in ErbB3-6F-expressing NIH-3T3 cells. 2 Together these results suggest that activation of Ras, a downstream target of the ErbB2/ErbB3 co-receptor (60) and a known activator of PI 3-kinase (22), does not always lead to PI 3-kinase activation. Our findings also indicate that activation of ErbB2 is not sufficient to activate PI 3-kinase signaling, at least in some cellular contexts. The results of these studies thus highlight the unique role of the ErbB3 protein in ErbB family receptor signaling.