Association of Atypical Protein Kinase C Isotypes with the Docker Protein FRS2 in Fibroblast Growth Factor Signaling*

FRS2 is a docker protein that recruits signaling proteins to the plasma membrane in fibroblast growth factor signal transduction. We report here that FRS2 was associated with PKC λ when Swiss 3T3 cells were stimulated with basic fibroblast growth factor. PKC ζ, the other member of the atypical PKC subfamily, could also bind FRS2. The association between FRS2 and PKC λ is likely to be direct as shown by yeast two-hybrid analysis. The C-terminal fragments of FRS2 (amino acid residues 300–508) and SNT2 (amino acids 281–492), an isoform bearing 50% identity to FRS2, interacted with PKC λ at a region (amino acids 240–562) that encompasses the catalytic domain.In vitro kinase assays revealed neither FRS2 nor SNT2 was a substrate of PKC λ or ζ. Mutation of the alanine residue (Ala-120) to glutamate in the pseudo-substrate region of PKC λ results in a constitutively active kinase that exhibited more than 2-fold greater binding to FRS2 in vitro than its “closed” wild-type counterpart. Tyrosine phosphorylation of FRS2 did not affect its binding to the constitutively active PKC λ mutant, suggesting that the activation of PKC λ is necessary and sufficient for its association with FRS2. It is likely that FRS2 serves as an anchoring protein for targeting activated atypical PKCs to the cell plasma membrane in signaling pathways.

FRS2 is a docker protein that recruits signaling proteins to the plasma membrane in fibroblast growth factor signal transduction. We report here that FRS2 was associated with PKC when Swiss 3T3 cells were stimulated with basic fibroblast growth factor. PKC , the other member of the atypical PKC subfamily, could also bind FRS2. The association between FRS2 and PKC is likely to be direct as shown by yeast two-hybrid analysis. The C-terminal fragments of FRS2 (amino acid residues 300 -508) and SNT2 (amino acids 281-492), an isoform bearing 50% identity to FRS2, interacted with PKC at a region (amino acids 240 -562) that encompasses the catalytic domain. In vitro kinase assays revealed neither FRS2 nor SNT2 was a substrate of PKC or . Mutation of the alanine residue (Ala-120) to glutamate in the pseudo-substrate region of PKC results in a constitutively active kinase that exhibited more than 2-fold greater binding to FRS2 in vitro than its "closed" wildtype counterpart. Tyrosine phosphorylation of FRS2 did not affect its binding to the constitutively active PKC mutant, suggesting that the activation of PKC is necessary and sufficient for its association with FRS2. It is likely that FRS2 serves as an anchoring protein for targeting activated atypical PKCs to the cell plasma membrane in signaling pathways.
Fibroblast growth factor receptors are members of the receptor-tyrosine kinase family (1). In contrast to other growth factor receptors such as those for EGF 1 and PDGF, FGF receptors are poorly auto-phosphorylated upon ligand binding. Instead, a 90-kDa protein called SNT1 or FGF receptor substrate-2 (FRS2) (2,3) is phosphorylated at multiple tyrosine sites. FRS2 has also been reported to be serine/threonine-phosphorylated in FGF-treated cell lysates (4). SNT2, a recently identified isoform of FRS2, has about 50% identity to FRS2 (5) mainly at the N-and C-terminal ends. FRS2 and SNT2 possess a myristoylation site and a PTB domain at the N terminus. The PTB domain is responsible for the direct interaction of FRS2 and SNT1 with the juxtamembrane region of the FGFR in a phosphotyrosine-independent manner. Deletion of the PTB domain of both proteins abrogates the association and tyrosine phosphorylation of FRS2 and SNT2 by FGF receptors (5,6).
FRS2 and SNT2 substitute for their receptors as docking proteins, a role similar to that of insulin receptor substrate (IRS) in insulin signaling (7). To date, two important signaling proteins, Grb2 and SHP-2, have been reported to bind directly to tyrosine-phosphorylated FRS2 (2,8). Grb2 is an adapter protein best known for its role in linking receptor tyrosine kinases to the Ras pathway via the guanine nucleotide-releasing factor Sos (9). The binding of Grb2 to FRS2 occurs via the interaction of the SH2 domain of Grb2 with some or all of the potentially phosphorylated tyrosine residues at Tyr-196, Tyr-306, Tyr-349, and Tyr-392 on FRS2. Mutational studies showed that when the tyrosine residues at all 4 sites were changed to phenylalanine, the downstream MAP kinase activation was significantly reduced (2). SHP-2 is a tyrosine phosphatase whose activity has been proposed to be necessary for cell growth and proliferation (10,11). When cells are stimulated with growth factors such as PDGF, SHP-2 is tyrosine-phosphorylated and binds to the SH2 domain of Grb2 (12). SHP-2 also binds to the activated receptors via its own SH2 domain (12). As a result, SHP-2 functions not only as a phosphatase but also serves as an adapter protein recruiting Grb2 to the receptors. Recently, SHP-2 has been reported to bind directly to tyrosinephosphorylated FRS2 through its N-terminal SH2 domain (8). The association of SHP-2 with FRS2 and the activation of SHP-2 are essential for a sustained MAP kinase response as well as for the potentiation of FGF-induced neurite outgrowth in PC12 cells (8). Hence, by recruiting Grb2 and SHP-2, FRS2 plays a crucial role in linking the FGF receptors to the Ras/ MAP kinase pathway.
Apart from the Grb2 and SHP-2 proteins, the activity of the atypical PKCs (aPKCs) is necessary for mitogenic signaling via the MAP kinase cascade (13,14). There are two members in the aPKC subfamily, PKC and PKC , and they share more than 75% identity. PKCs have been subdivided into 3 subfamilies, and they are distinguished by their lipid activation profiles. Conventional PKCs (cPKCs e.g. ␣, ␤, and ␥) are activated by diacylglycerol and calcium; novel PKCs (nPKCs e.g. ␦, ⑀, , and ) do not respond to calcium but require diacylglycerol for their activation; and aPKCs are not activated by either diacylglycerol or calcium. It has been shown that MAP kinase and MEK are activated in vivo by an active mutant of PKC , and a kinase-defective dominant negative mutant of PKC impairs the activation of both MEK and MAP kinase by serum and tumor necrosis factor (14). However, whereas Grb2 and SHP-2 lie upstream of Ras, PKC can bind to and act as a direct effector of Ras (15). This is consistent with the observation that expression of a dominant negative mutant of Ras (Asn-17) severely impairs the activation of PKC by mitogens such as PDGF in mouse fibroblasts (15).
A few groups of proteins that are either regulators or substrates of aPKCs bind to the members in this subfamily. In the first group, a protein called Zeta-Interacting Protein (ZIP) binds specifically to the regulatory domain of PKC (16), whereas Lambda-Interacting Protein (LIP) binds specifically to the regulatory domain of PKC resulting in an activation of the kinase (17). The Par-4 protein also binds to the regulatory domain of PKC and PKC but inhibits their activity (18). The second group comprises proteins like heterogeneous ribonucleoprotein A1 protein that has been found to bind to the kinase domain of PKC in yeast two-hybrid screening and is a specific substrate of the aPKCs (19).
We have been studying p75, a phosphotyrosine protein that is dephosphorylated and dissociates from Grb2 upon growth factor stimulation (27). In our attempt to identify p75 by immunoprecipitating phosphotyrosyl proteins that are about 75-kDa in molecular mass, we observed that a 90-kDa tyrosinephosphorylated protein, p90, associates specifically with members of the aPKC subfamily but not with other PKC family members. In this report, we identified the p90 protein as FRS2. We have also characterized the factors that regulate its association with the aPKCs. We propose that FRS2 plays an important role in the targeting of activated PKC or to the plasma membrane. Thus FRS2 may constitute a third group of proteins that bind to the aPKCs and localize them in specific subcellular compartments. The recruitment of aPKCs by FRS2 to the cell-surface membrane may be an important event contributing to the regulation of the aPKC activity.

EXPERIMENTAL PROCEDURES
Reagents-Monoclonal antibodies against phosphotyrosine (PY20), Grb2, SHP-2, and PKC were purchased from Transduction Laboratories (Lexington, KY). Polyclonal antibodies against PKC ␣, ␦, and PKC / were from Santa Cruz Biotechnology (Santa Cruz, CA). Polyclonal antibodies against FRS2 (A872) were raised against amino acids (residues 491-506) and produced by Neosystem Laboratoire (Strasbourg, France). Secondary anti-mouse and anti-rabbit antibodies conjugated to horseradish peroxidase were from Sigma, and protein A/G plus agarose was from Santa Cruz Biotechnology. Anti-activation domain and antibinding domain antibodies are from CLONTECH (Palo Alto, CA). Recombinant human EGF and PDGF were from Sigma, and basic FGF (bFGF) was from Roche Molecular Biochemicals (Mannheim, FRG). PKC , PKC ␣, PKC ␦, and PKA purified enzymes were from Life Technologies, Inc.
Cell Lines, Cell Stimulation, and Lysis-Swiss 3T3 fibroblasts (ATCC CCL92, Rockville, MD) were grown and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (HyClone Laboratories, Logan, UT), 2 mM glutamine, 10 mM HEPES, pH 7.4, and 100 units/ml penicillin and streptomycin. Human 293T kidney epithelial cells were grown in 150-mm culture dishes with RPMI medium supplemented with 10% fetal bovine serum (HyClone Laboratories), 2 mM glutamine, 10 mM HEPES, pH 7.4, and 100 units/ml penicillin and streptomycin. When the cells were about 80 -90% confluent, the medium was aspirated, and the cells were washed and maintained for another 18 -24 h in serum-free medium. Various growth factors were added to the quiescent cells prior to aspiration of the medium. The cells were then washed rapidly in cold phosphatebuffered saline and lysed in 500 l of lysis buffer containing 0.5% Nonidet P-40, 20 mM Tris-HCl, pH 7.3, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 10% glycerol, 1 mM sodium orthovanadate, and a mixture of protease inhibitors (Roche Molecular Biochemicals) added according to the manufacturer's instructions. The cell lysates were spun at 11,000 ϫ g for 5 min at 4°C, and the supernatants were used for subsequent analyses. The protein concentrations of all cell lysates were normalized after estimation of their protein content using a BCA protein assay kit from Pierce.
Construction of Plasmids-PKC / cDNA and HA-tagged PKC in pCDNA3 were kind gifts from Dr. Jorge Moscat (Universidad Autonoma de Madrid, Spain). PKC ␤II cDNA and PKC ␦ cDNAs were from Dr. Alexandra Newton (University of California, San Diego) and Dr. Li Weiqun (National Cancer Institute, Bethesda), respectively. cDNAs encoding the full-length PKC , PKC fragment A (amino acids 1-239), PKC fragment B (amino acids 240 -586), PKC fragment B (amino acids 240 -592), PKC ␦ fragment B (amino acids 354 -701), and PKC ␤II fragment B (amino acids 345-673) were obtained by PCR. These inserts were introduced into pGEX4T1 vector for the expression of GST fusion proteins in bacterial cells. FRS2 cDNA was obtained first by reverse transcription from mRNA extracted from Swiss 3T3 cells. PCR was then carried out with the following primers, which were designed based on the published sequence of FRS2 (2), to obtain the full-length cDNA as follows: (forward) 5Ј CGC GGA TCC GCG ATG GGT AGC TGT TGT AGC TGT CC 3Ј and (reverse) 5Ј CG GCGG CCGC TCA CAT GGG CAG GTC AGT ACT ATT G 3Ј. The BamHI/NotI insert was introduced into pGEX4T1 and pXJ40HA for the expression of GST fusion protein in bacteria cells and HA-tagged proteins in mammalian cells, respectively. The expressed proteins were partially microsequenced and shown to be authentic. The FRS2 fragments X (amino acids 1-152), Y (amino acids 153-300), Z (amino acids 301-508), XY (amino acids 1-300), and YZ (amino acids 153-508) were obtained by PCR using the full-length FRS2 cDNA as template. All the inserts were cloned into pGEX4T1 for the expression of GST fusion proteins. Human SNT2 cDNA was a kind gift from Dr. Mitchell Goldfarb (Mount Sinai School of Medicine, New York). cDNA encoding the fragment Z (amino acids 281-492) of SNT2 was obtained by PCR and cloned into pGEX4T1 for the production of GST fusion proteins. The cDNA for human FGF receptor 1 (Flg) was a kind gift from Dr. Lena Claesson-Welsh (Ludwig Institute for Cancer Research, Uppsala, Sweden). cDNA encoding the cytoplasmic domain (amino acids 398 -822) of Flg was obtained by PCR and cloned into pXJ40Flag for the expression of Flag-tagged cytosolic Flg in mammalian cells.
For yeast two-hybrid screening, cDNAs encoding the full-length fragment A (amino acids 1-239) and fragment B (amino acids 240 -586) of PKC were obtained by PCR as described above and introduced into pAS vector suitable for yeast transformation and expression of Gal4binding domain fusion protein. SHP-2 cDNA was a kind gift from Dr. Tony Pawson (Mount Sinai Hospital, Ontario, Canada). Full-length FRS2 and SHP-2 were subcloned into pACT vector for yeast expression of Gal4 activation domain fusion proteins. cDNAs for the tandem SH2 domains (amino acids 1-213) and PTP catalytic domain (amino acids 214 -603) of SHP-2 were obtained via PCR, and the inserts were also cloned into the pACT vector.
Mutagenesis-Mutation of alanine to glutamate A120E in the pseudo-substrate site of PKC was carried out using the QuickChange mutagenesis kit from Stratagene (La Jolla, CA) according to the manufacturer's instruction. The template used was wild-type full-length PKC in pGEX4T1 and pXJ40HA. The primers used were as follows 5Ј CCG GAG AGG GGA ACG CCG TGG GAG 3Ј and 5Ј CTC CAC CGG CGT TCC CCT CTC CGG 3Ј. The products were sequenced and verified to be correct.
Transfections-Human 293T kidney epithelial cells were grown in 100-mm culture dishes as described above. Cells that were about 90% confluent were used for transfection. For single or co-transfections, 15 g of each DNA followed by 4.5 l/g DNA of TfX 50 from Promega (Madison, WI) were added to 6 ml of serum-free RPMI and incubated at room temperature for 15 min. The transfection mix was then added to cells prewashed with serum-free medium and left at 37°C for 1 h. After this, 12 ml of RPMI supplemented with 10% fetal bovine serum was added, and the cells were left to recover for 48 h. The cells were lysed with RIPA buffer (50 mM Tris-HCl, pH 7.3, 150 mM NaCl, 0.25 mM EDTA, 1% sodium deoxycholate, 1% Triton X-100, 1 mM sodium orthovanadate, and a mixture of protease inhibitors from Roche Molecular Biochemicals) and processed as described under "Cell Lines, Cell Stimulation, and Cell Lysis." GST Fusion Proteins-All the constructs for the production of GST fusion proteins were transformed into DH5␣ cells. The transformed cells were grown in 1 liter of LB ϩ ampicillin (50 g/ml) medium and incubated at 37°C with shaking (220 rpm) to an A 600 of about 0.3. These cells were then induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside at room temperature overnight. The cells were spun down and frozen at Ϫ80°C. The cell pellet was then left to thaw on ice, and 10 ml of lysis buffer (phosphate-buffered saline, 1% Triton X-100, 1 mM dithiothreitol, and a mixture of protease inhibitors from Roche Molecular Biochemicals) was added to the cell. The cell suspension was subse-quently sonicated for a total of 12 pulses of 15 s with a 30-s pause between each pulse. The lysates were centrifuged and supernatants were incubated with glutathione beads overnight at 4°C to purify the GST fusion proteins.
Immunoprecipitations, in Vitro Binding Assays-Quiescent or activated cells were lysed as described above, and an equal volume of 2ϫ precipitation buffer (20 mM Tris-HCl, pH 7.4, 300 mM NaCl, 2% Triton X-100, 2 mM EDTA, 2 mM EGTA, and 1% Nonidet P-40) was added to the cell lysate. For immunoprecipitation, 2.5 g of the appropriate antibodies were added to the diluted cell lysate and incubated for 1 h or overnight at 4°C. 2.5 g of secondary antibodies conjugated to agarose was added to capture the immunocomplex for 1 h or overnight at 4°C. In the depletion studies, the immunoprecipitation described above was repeated 5 times, each for 1.5 h. After washing, the immunoprecipitates were pooled together and resolved by SDS-PAGE.
For in vitro binding assays with GST fusion proteins, 10 g of the GST fusion proteins were incubated with the lysates for 1 h or overnight at 4°C. All the beads were washed 3 times with 1ϫ precipitation buffer, and the bound proteins were eluted with 2ϫ Laemmli buffer before separation by SDS-PAGE.
Yeast Two-hybrid Analysis-The various constructs of PKC (full length, fragment A, and fragment B) in pAS were sequenced and verified to be correct before they were introduced into yeast strain 190 using the yeast transformation kit from CLONTECH (Palo Alto, CA). The transformed yeast were grown in selective media SD-Trp at 30°C until colonies appeared. Single transformants then underwent a second round of transformation with the pACT vectors containing full-length FRS2, full-length SHP-2, SH2 domains, or PTP catalytic domain of SHP-2. Successful dual transformants were selected on SD-Trp/Leu, and LacZ blue assays were carried out, according to the manufacturer's instructions, to detect for protein interactions in the yeast. Yeast transformed with pCL1 expressing functional Gal4 protein turned blue between 0.5 and 1 h. This serves as a positive control for the LacZ assay. Colonies turning blue after 8 h were considered negatives according to the manufacturer's instruction. The dual transformants were also analyzed for the expression of the various proteins by first growing them in SD-Trp/Leu liquid medium. The yeast was lysed according to the transformation kit manufacturer's instructions, and the lysates were separated on SDS-PAGE. Following Western blotting, the various proteins were detected by probing with the appropriate antibodies.
In Vitro Kinase Assay-For activation studies of aPKCs by growth factors, Swiss 3T3 cells were either untreated or stimulated with bFGF at 20 ng/ml for 10 min. After the cells were lysed, immunoprecipitations of aPKC and subsequent in vitro kinase assays were carried out as described elsewhere (4). In vitro kinase assays were also carried out either with purified PKA, PKC , PKC ␣, or PKC ␦ enzyme purchased from Life Technologies, Inc. HA-tagged A120E PKC mutant and HA-tagged PKC fragment B (containing the kinase domain) were also used as a source of kinase activities. In cases where the HA-tagged PKC proteins were used, immunoprecipitations using HA antibodies were carried out before the kinase assays were performed. GST full-length FRS2 or GST-FRS2 fragment Z was tested as substrate for aPKCs, and hnRNPA1, a gift from Dr. Jorge Moscat (Universidad Autonoma de Madrid, Spain), and MBP were used as positive controls. All aPKCs reactions were carried out at 30°C for 30 min in 20 l of buffer (35 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 0.5 mM EGTA, 0.1 mM CaCl 2 , and 1 mM phenylphosphate) containing 50 ng of enzymes, 2 g of substrate, 5 Ci of [␥-32 P]ATP (Amersham Pharmacia Biotech), and 50 M of ATP. The reactions for the PKC ␣ and PKC ␦ were carried out in 20 l of buffer (20 mM HEPES, pH 7.4, 1.5 mM CaCl 2 , 1 mM dithiothreitol, and 10 mM MgCl 2 ) containing 50 ng of enzymes, 50 g/ml sonicated phosphatidylserine, 2 g of substrates, 5 Ci of [␥-32 P]ATP (Amersham Pharmacia Biotech), and 50 M of ATP. The reaction was stopped by boiling with an equal volume of 2ϫ Laemmli buffer, and the proteins were separated on SDS-PAGE. The gel was dried and subjected to autoradiography.

RESULTS
FRS2 Co-precipitates with PKC in Response to FGF Stimulation-Our laboratory has been characterizing p75 and its association with Grb2. In quiescent cells, p75 is tyrosine-phosphorylated and binds to the SH2 domain of Grb2. Upon stimulation with growth factors including FGF, p75 is dephosphorylated and dissociates from Grb2 (27). We were keen to identify p75 and decided to test existing phosphotyrosyl proteins that are about 75 kDa for dephosphorylation upon FGF treatment. One of the candidates that we selected was PKC ␦, a member of the nPKC subfamily which is about 78 kDa and is the only PKC member currently known to be tyrosine-phosphorylated (33). We therefore set out to investigate whether PKC ␦ is tyrosine-phosphorylated in quiescent cells. Representative members, namely PKC ␣ and PKC , from the cPKC and aPKC subfamilies, respectively, were also included for comparison.
Preliminary studies in our laboratory have shown that two FGF-responsive cell lines, Swiss 3T3 and PC12 cells, expressed all the three PKCs of interest. Swiss 3T3 cells were chosen for further experiments because they respond better to bFGF than PC12 cells. To examine whether PKC ␦, PKC ␣, or PKC could be the p75 that undergoes dephosphorylation upon growth factor stimulation, immunoprecipitation of the various PKCs was carried out on lysates on Swiss 3T3 cells that were either untreated or stimulated with bFGF. The immunoprecipitates were resolved by SDS-PAGE and Western blotted. The membrane was probed with phosphotyrosine antibodies to detect for the presence of tyrosine-phosphorylated PKCs. None of the PKCs was tyrosine-phosphorylated in the lysates of quiescent cells (Fig. 1A, upper panel). The blot was stripped and reprobed either with PKC ␣, PKC ␦, or PKC antibodies to show that the individual PKCs had been immunoprecipitated (Fig.  1A, lower panel). We conclude that none of the PKCs tested are likely to be p75.
In these experiments, we noted that a 90-kDa tyrosine-phosphorylated protein, similar to an FGF-specific p90 protein that has been reported to bind Grb2 (34), was co-immunoprecipitated with PKC (Fig. 1A, upper panel). Neither PKC ␦ nor PKC ␣ co-precipitated this tyrosine-phosphorylated protein significantly when compared with PKC in the lysates from bFGF-stimulated cells. We therefore decided to investigate this apparently specific association.
It is possible that in the experiments carried out above, differential amounts of the various PKCs were immunoprecipitated by the antibodies due to the different affinities of the individual antibodies for their respective PKCs. The apparently larger amount of p90 co-immunoprecipitated with PKC may due to higher amounts of PKC immunoprecipitated compared with the other PKCs. Hence, it was necessary to ensure that the majority (Ͼ80%) of each PKC was immunoprecipitated. Preliminary optimization showed that five successive rounds of immunoprecipitation were enough to deplete 80% or more of the various PKCs (data not shown). Therefore, five rounds of immunoprecipitation of PKC , PKC ␣, and PKC ␦ (as described under "Experimental Procedures") were carried out on lysates from Swiss 3T3 cells that have been stimulated with bFGF. The immunoprecipitates were pooled, resolved by SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. The membrane was then probed with phosphotyrosine antibodies to detect p90. Fig. 1B, top panel, shows that p90 co-immunoprecipitated only with PKC . The blot was stripped and cut between lanes and probed for the various PKCs immunoprecipitated (Fig. 1B, middle panel). The amounts of the various PKCs present in the lysates before and after multiple immunoprecipitations were assessed by Western blot analyses. Fig. 1B, bottom panel, shows that more than 80% of PKC , PKC ␣, or PKC ␦ were immunoprecipitated. Therefore, the co-immunoprecipitation of p90 with PKC is not due to a relatively larger proportion of PKC being immunoprecipitated compared with PKC ␣ or PKC ␦.
The molecular mass and the gel migration pattern of the 90-kDa tyrosine-phosphorylated protein resembled that of FRS2, a protein that our laboratory is currently studying. To determine whether the p90 protein was FRS2, lysates from Swiss 3T3 cells that were untreated or stimulated with bFGF were subjected to immunoprecipitation of PKC . The immu-noprecipitates were processed as described above. The blot was first probed with phosphotyrosine antibodies revealing the 90-kDa tyrosine-phosphorylated protein co-precipitating with PKC upon bFGF stimulation (Fig. 1C, top panel). The blots were stripped and re-probed with A872, a polyclonal antibody raised against FRS2. As shown in Fig. 1C, middle panel, FRS2 co-precipitated with PKC from lysates of bFGF-stimulated but not non-stimulated cells. It is noted that phosphotyrosine signal for p90 that co-immunoprecipitated with PKC (Fig. 1C, top panel) is greater than that of FRS2 co-immunoprecipitated with PKC (Fig. 1C, middle panel). This is attributed to the observation that FRS2 is a multiply tyrosine-phosphorylated protein with at least 6 tyrosine phosphorylation sites. By comparing the amount of FRS2 in the total lysate with the amount co-immunoprecipitated with PKC in other independent experiments (data not shown), it is estimated that the amount of FRS2 co-immunoprecipitated with PKC was approximately 5%. Probing this blot with PKC antibodies showed equal amounts of PKC being immunoprecipitated from both lysates (Fig. 1C, bottom panel). The experiment in Fig. 1C has also been repeated by probing the blot first with FRS2 antibodies followed by anti-phosphotyrosine antibodies. Similar results were obtained (data not shown). This demonstrated that FRS2 and PKC exist in a complex following bFGF stimulation of Swiss 3T3 cells. The association of FRS2 with PKC in FGFstimulated cells might be 1) mediated by other proteins, 2) dependent on the activation of PKC , 3) dependent on the tyrosine phosphorylation of FRS2, or 4) a combination of some or all the above factors. We therefore set out to address these possibilities.
FRS2 Binds to the Catalytic Domain of PKC -Although FRS2 co-immunoprecipitated with PKC , it is possible that the association of PKC with FRS2 is mediated through other proteins in the immunoprecipitated complex. We have shown previously that SHP-2 associates with FRS2, and it is possible that PKC associates directly with SHP-2 and not FRS2. Grb2 also binds to FRS2 but experiments showed that Grb2 was not present in complexes containing FRS2 and PKC (data not shown). We employed the yeast two-hybrid technique to investigate whether FRS2 is likely to bind to PKC directly. The PKC protein exists in a "closed," catalytically inactive state in non-stimulated cells. In addition to full-length PKC , two other fragments that contained the regulatory domain (fragment A) or the catalytic domain (fragment B) of PKC were generated. These fragments would theoretically "expose" potential regions in the protein that are normally masked. PKC fragment A contains amino acids 1-239 and encompasses the N terminus extension, pseudo-substrate site, zinc finger, and part of the hinge region; fragment B contains amino acids 240 -586 and includes part of the hinge, the kinase domain, and carboxyl tail. Plasmids containing these proteins were transformed into yeast as described under "Experimental Procedures." To study the interaction of FRS2 with the various fragments of PKC , yeast expressing the various PKC proteins underwent another transformation with an expression vector encoding fulllength FRS2. In addition, the SHP-2 full-length, the tandem Nand C-terminal SH2 domains, and the PTP catalytic domain of SHP-2 were also separately introduced to assess their binding various PKCs. C, association of FRS2 with PKC in Swiss 3T3 cells. Serum-deprived cells were either non-stimulated or stimulated with 10 ng/ml bFGF and the lysates subjected to immunoprecipitation with PKC antibodies. Immunoprecipitates were separated on SDS-PAGE and Western blotted. Top panel, the blot was first probed with PY20 phosphotyrosine antibodies. Middle panel, the blot was stripped and probed with FRS2 antibodies. Bottom panel, the blot was stripped a third time and probed with PKC antibodies. FIG. 1. A, co-immunoprecipitation of a tyrosine-phosphorylated 90-kDa protein with PKC . Quiescent Swiss 3T3 cells were either not stimulated or stimulated with 10 ng/ml bFGF for 10 min. The cells were lysed and the lysates subjected to immunoprecipitation (IP) of PKC , PKC ␦, or PKC ␣ as described under "Experimental Procedures." Upper panel, the immunoprecipitates were separated by SDS-PAGE and immunoblotted (IB) with phosphotyrosine antibodies (PY20). The arrows indicate the positions of p90 and p75. Lower panel, the blot was stripped and re-probed with PKC , PKC ␦, or PKC ␣ antibodies to reveal the amount of the various PKCs immunoprecipitated. B, multiple immunoprecipitations of PKC , PKC ␣, and PKC ␦. Swiss 3T3 cells were stimulated with bFGF (10 ng/ml) for 10 min and lysed. The lysates were subjected to 5 successive rounds of immunoprecipitation of PKC , PKC ␦, and PKC ␣ as described under "Experimental Procedures" before the immunoprecipitates for each PKC were pooled, resolved by SDS-PAGE, and Western blotted. Top panel, the membrane was probed with phosphotyrosine antibodies. Middle panel, the blot containing the immunoprecipitates was cut into strips and re-probed either with PKC , PKC ␣, or PKC ␦ antibodies. These individual strips were then re-aligned and the respective PKCs detected using the ECL reagents to reveal the proteins immunoprecipitated. Bottom panel, the immunoprecipitation efficiency for each of the PKCs was assessed by probing the lysates before (Pre) or after (Post) successive rounds of immunoprecipitation with the respective antibodies. The arrows indicate the positions of the to PKC . In all cases, colonies expressing all combination of proteins were obtained (data not shown) and were subjected to LacZ assays. Yeast transformed with pCL1 expressing functional Gal4 protein turned blue between 0.5 and 1 h. This serves as a positive control for the LacZ assay. Colonies turning blue after 8 h were considered negatives according to the manufacturer's instruction. In the above assays, only yeast expressing FRS2 and PKC fragment B (encompassing the kinase domain) turned blue (at 2.5 h), indicating a strong interaction (Table I). Of particular note was the observation that fulllength PKC and fragment A containing the PKC regulatory domain did not interact with FRS2. Thus, FRS2 interacts with PKC through a region (amino acids 240 -586) that is predominantly the catalytic domain. This region is probably masked in the non-activated, full-length molecule since the latter cannot interact with FRS2. The observation that the full-length SHP-2, the tandem SH2 domains, and the PTP catalytic domains of SHP-2 did not interact with any of the PKC polypeptides indicates that the interaction between FRS2 and PKC is both specific and most likely direct. FRS2 is not likely to be an activator of PKC since the well characterized activators of the kinase bind to the regulatory region of the kinase (16,17).
To verify the yeast two-hybrid results, GST-PKC fragment A and GST-PKC fragment B were tested for their ability to bind to FRS2. Since FRS2 co-immunoprecipitated with PKC only upon FGF stimulation, Swiss 3T3 cells were stimulated with bFGF and the lysates incubated with either GST-PKC fragment A or fragment B. An aliquot of the lysate was also incubated with GST as a control. The precipitates were resolved by SDS-PAGE and Western blotted. The presence of FRS2 in the precipitates was detected by probing the membrane with phosphotyrosine antibodies rather than FRS2 antibodies because of the ease of detection as well as the avoidance of the high nonspecific background signal encountered when using FRS2 antibodies. Consistent with the yeast twohybrid results, Fig. 2A shows that fragment B of PKC is responsible for binding to FRS2. The amounts of tyrosinephosphorylated FRS2 seen to bind to PKC fragment A and GST represent background signal.
To determine whether the binding of FRS2 to PKC fragment B is specific, various GST fusion proteins were prepared for in vitro binding assays. Lysates of bFGF-stimulated Swiss 3T3 cells were incubated with the fragment B of PKC or, for comparison, fragment B of PKC ␤II (a member of the cPKC family) and PKC (the other member of the aPKC). The pre-cipitates were resolved by SDS-PAGE and immunoblotted with phosphotyrosine antibodies to detect for tyrosine-phosphorylated FRS2. Fig. 2B, top panel, shows that the fragment B of PKC but not that of PKC ␤II precipitated tyrosine-phosphorylated FRS2. Fragment B of PKC , the other member of the atypical PKCs (aPKCs), also bound tyrosine-phosphorylated FRS2. This is not surprising given the observation that the fragment B containing the kinase domains of PKC have more than 85% identity with PKC . On the other hand, the kinase domain of PKC ␤II did not bind significant amounts of FRS2 compared with the aPKCs. The fragment B of PKC ␦ also did not bind FRS2 significantly (data not shown). This reflects a lack of affinity in the fragment B of cPKCs and nPKCs for FRS2. Together, the above results demonstrate that only the members of the atypical PKC subfamily interact with FRS2 through a region that encompasses the catalytic domain, and this interaction is most likely to be direct.
PKC Binds to a Region in the C Terminus of FRS2-To define the region on FRS2 that binds to PKC , GST fusion proteins containing various fragments of FRS2 were produced (Fig. 3A). Lysates of human 293T cells grown in 10% serum were incubated with each of the GST-FRS2 polypeptides overnight at 4°C. The precipitates were washed and resolved by SDS-PAGE and immunoblotted with PKC monoclonal antibodies (Fig. 3B). Fragment Z (amino acids 301-508) of FRS2 precipitated a comparatively larger amount of PKC compared with fragments X and Y. Consistent with this observation, full-length FRS2 and fragment YZ but not fragment XY could also bind PKC . However, full-length FRS2 and FRS2 fragment YZ bind slightly lesser PKC than fragment Z, suggest- two-hybrid assay Following co-transformation of yeast strain Y190 with a combination of plasmids shown in the table, dual transformants were selected on SD-Trp/Leu-selective media. Colonies that grew were subjected to LacZ assay as described under "Experimental Procedures" to test for the induction of ␤-galactosidase activity that results from interaction between two proteins expressed in the yeast. Blue colonies indicate a positive interaction, and white indicates a negative interaction. ing that other parts of FRS2 may interfere with the binding of PKC . It is also possible that the GST-FRS2 full-length or fragment YZ may not have been folded properly during the preparation. As a control, the GST-FRS2 fragment Z alone was resolved by SDS-PAGE, Western blotted, and probed with PKC antibodies. No signal was obtained (data not shown) indicating that the PKC signal from fragment Z in Fig. 3B was not intrinsic to the GST fusion proteins.
To define further the binding region on FRS2, fragment Z of FRS2 was subdivided into smaller fragments, and in vitro binding assays similar to those described above were per-formed. PKC showed essentially equal binding to each of the subfragments of fragment Z (data not shown), indicating that binding occurs at multiple points within this peptide sequence.
Next, we investigated whether SNT2, an isoform that has 50% identity to FRS2, can bind to PKC . A GST fusion protein of SNT2 fragment Z (amino acids 281-492), whose sequence was aligned with that of FRS2 fragment Z, was produced and assessed for its binding to PKC . The fragment Z from FRS2 and SNT2 share about 50% identity. GST full-length FRS2 and GST fragment Z of FRS2 were included for comparison, and a similar experiment to that described above was carried out. Fig. 3C shows that fragment Z of SNT2 could bind endogenous PKC as well as fragment Z from FRS2.
"Activated" PKC Binds to FRS2-By having demonstrated that the fragment B of PKC binds to the carboxyl portion of FRS2, we proceeded to investigate whether activation of PKC was necessary for its association with FRS2. First, we examined the activation status of PKC by FGF. Swiss 3T3 cells were either not stimulated or stimulated with bFGF. The cells were lysed, and immunoprecipitation of PKC was performed on these lysates. Subsequently, the immunoprecipitated PKC was tested for in vitro kinase activity toward hnRNPA. hnRNPA1 is the only convincing aPKC substrate identified so far (19). Fig. 4A showed that bFGF stimulated the activity of PKC by about 3-fold. This level of activation is comparable to that obtained by stimulating NIH3T3 cells with PDGF (19). The opening of the otherwise closed PKC protein as a result of activation by FGF may expose sites that are necessary for its association with FRS2. This is consistent with the yeast twohybrid results (Table I) where only fragment B of PKC binds to FRS2. Apparently, the masking of this fragment in the inactive, full-length PKC contributed to its inability to bind FRS2.
We postulated that if the inactive PKC could be "artificially" opened up, thus exposing the sites represented on fragment B, full-length PKC might acquire the ability to bind FRS2. It has been shown previously that mutation of the alanine residue in the pseudo-substrate site of PKC to glutamate would switch the kinase to a constitutively active form (21,22). We therefore set out to investigate the relative affinities of the constitutively active PKC A120E mutant and the wild-type PKC for tyrosine-phosphorylated FRS2. Although the data from yeast two-hybrid analysis showed that non-tyrosine-phosphorylated FRS2 can bind to fragment B of PKC , tyrosine-phosphorylated FRS2 was used in the assay because tyrosine phosphorylation may enhance the binding to PKC (refer to Fig. 1C). To eliminate any potential contribution of SNT2, we transfected 293T cells with HA-tagged FRS2 (HA-FRS2). We have previously noted that the overexpression of the cytosolic fragment of FGFR1 (Flg) leads to the activation of the endogenous tyrosine kinase activity without the addition of exogenous bFGF (data not shown). Hence, to obtain tyrosine-phosphorylated HA-FRS2, 293T cells were co-transfected with the cytosolic fragment of Flg (Flg-cyto) and HA-FRS2. Total cell lysates were separated on SDS-PAGE and Western blotted. Probing the blot with FRS2 (A872) antibodies showed that the level of HA-FRS2 was much higher than the endogenous FRS2, which was present in very low abundance (data not shown). Immunoprecipitation with HA antibodies showed that the expressed HA-FRS2 was tyrosine-phosphorylated (data not shown).
To assess the binding of the constitutively active PKC mutant or wild-type PKC to tyrosine-phosphorylated FRS2, 293T cells were co-transfected with HA-FRS2 and the cytosolic fragment of Flg. The cell lysates containing tyrosine-phosphorylated HA-FRS2 were then incubated with equal amounts of GST-PKC A120E mutant or GST-wild-type PKC . The GST FIG. 3. A, diagram showing the various GST fusion proteins of FRS2 produced. The tyrosine residues Y196, Y306, Y349, and Y392 are potential Grb2-binding sites and Y436 and Y471 are potential SHP-2binding sites. B, PKC binds to a C-terminal region of FRS2. 293T cells were grown in 10% serum and lysed when 95% confluent. The lysates were incubated with 10 g of GST fusion proteins of FRS2 full-length (FL), fragment X, fragment Y, fragment Z, fragment XY, and fragment YZ as described under "Experimental Procedures." The precipitates were washed, eluted, and the proteins separated on SDS-PAGE. After the Western blotting, the membrane was probed with PKC antibodies. C, PKC also binds to a C-terminal region of SNT2. 293T cells were grown in 10% serum and lysed when they were 95% confluent. The lysates were incubated with 10 g of the GST fusion proteins of fulllength FRS2 (FL), FRS2 fragment Z, and SNT2 fragment Z. The precipitates were processed as described in B, and the presence of PKC was examined by immunoblotting (IB) with PKC antibodies. PD, pull-down. fusion proteins containing the fragment B of PKC or PKC ␤II were also included for comparison. The precipitates were separated by SDS-PAGE, Western blotted, and the membrane probed with phosphotyrosine antibodies. Fig. 4B revealed that the constitutively active PKC mutant binds much more tyrosine-phosphorylated FRS2 protein than its wild-type counterpart. The amount of FRS2 bound by wild-type PKC was about the same as that bound by PKC ␤II and is likely to represent the background level of binding as we have shown previously that FRS2 does not associate with this PKC isoform (Fig. 2B). In addition, the PKC fragment B is able to bind as much FRS2 as PKC A120E mutant indicating that fragment B of the PKC molecule is sufficient for binding FRS2 without the cooperation of other parts of the molecule. A similar experiment was also carried out using lysates from Swiss 3T3 cells that were stimulated with FGF. The lysates were incubated with GST-PKC A120E mutant or GST-wild-type PKC , and the precipitates were probed with phosphotyrosine antibody to detect endogenous tyrosine-phosphorylated FRS2. The results (data not shown) were essentially the same as Fig. 4B.
The in vitro binding of mutant and wild-type PKC to FRS2 was also assessed using GST-FRS2 fragment Z, and the lysates of 293T cells were transfected with HA-tagged wild-type or constitutively active mutant PKC . The transfected cells expressed equivalent amounts of HA-tagged mutant and wildtype PKC (Fig. 4C, top panel). Following incubation of the cell lysates with GST-FRS2 fragment Z, the precipitated proteins were separated on SDS-PAGE, Western blotted, and probed with HA antibodies. As shown in Fig. 4C (bottom panel), more mutant compared with the wild-type PKC was bound to GST-FRS2 fragment Z. This strengthens the observation that constitutively active PKC A120E mutant has a stronger affinity for FRS2 than the inactive wild-type PKC and indicates that activation of PKC is required to bind to FRS2.
Tyrosine Phosphorylation of FRS2 Is Not Required for Association with aPKC-Although the activation of PKC can account for the induced association of the two proteins, we cannot exclude the possibility that tyrosine phosphorylation of FRS2 is also a factor regulating this association. Therefore, we set out to investigate the role of tyrosine phosphorylation in the FRS2/PKC interaction. 293T cells were transfected with either HA-FRS2 alone or with cytosolic Flg, allowing expression of nonphosphorylated or tyrosine-phosphorylated HA-FRS2. The cells were lysed and the lysates incubated with equal amounts of GST fusion proteins containing the constitutively active PKC mutant or wild-type PKC . An equivalent amount of agarose-conjugated GST protein was included as a control. The precipitates were separated by SDS-PAGE and immunoblotted with phosphotyrosine antibodies to detect FRS2. As expected, tyrosine-phosphorylated HA-FRS2 was precipitated by mutant PKC . Very low levels of tyrosinephosphorylated FRS2 were seen with wild-type PKC and GST alone, and these represented background signals (Fig. 5, top A120E mutant than the HA-tagged wild-type PKC . Upper panel, 293T cells were transfected with either HA-tagged PKC A120E mutant or wild-type PKC or were not transfected. The cells were lysed after 48 h of recovery, the lysates separated on SDS-PAGE, and Western blotted. The blot was then probed with HA antibodies to reveal the expression of the various tagged proteins. Lower panel, the lysates from 293T cells transfected with either HA-tagged PKC mutant or wild-type PKC were incubated with 10 g of the GST-FRS2 fragment Z as described under "Experimental Procedures." The precipitates were washed, eluted, and separated on SDS-PAGE. After Western blotting, the blot was probed with HA antibodies to reveal the amount of HA-tagged proteins that were precipitated. The arrow indicates the positions of the HA-tagged wild-type PKC or PKC A120E mutant. IB, immunoblotting; PD, pull-down.

FIG. 4.
A, PKC is activated by bFGF in Swiss 3T3 cells. Swiss 3T3 cells were either non-stimulated or stimulated with bFGF at 20 ng/ml for 10 min before the cells were lysed. The lysates were subjected to immunoprecipitation using PKC antibodies. After the immunoprecipitates were washed, they were used for in vitro kinase assays using eluted hnRNPA1 as substrate as described under "Experimental Procedures." After the reaction, the proteins were resolved by SDS-PAGE. The gel was dried and exposed to x-ray film (Fuji). The phosphorylated protein bands obtained were quantitated using a densitometer (Bio-Rad), and the relative activity of PKC is shown. The values shown represent the average Ϯ half the range. B, constitutively active PKC A120E mutant binds more HA-tagged FRS2 than wild-type inactive PKC . 293T cells were co-transfected with HA-tagged FRS2 (HA-FRS2), and the cytosolic domain of Flg (Flg-cyto). The cells were lysed after 48 h of recovery and the lysates incubated with 10 g of GST fusion proteins of PKC A120E mutant (mt), wild-type (wt) PKC , PKC ␤II fragment B, and PKC fragment B as described under "Experimental Procedures." The precipitates were washed, eluted, and separated on SDS-PAGE. After Western blotting the blot was probed with phosphotyrosine antibodies to reveal the amounts of tyrosinephosphorylated FRS2 precipitated by the various fusion proteins. C, GST-FRS2 fragment Z binds with a higher affinity to HA-tagged PKC panel). When the blot was stripped and re-probed with FRS2 (A872) antibodies (Fig. 5, bottom panel), equivalent amounts of FRS2 were shown to be precipitated by the PKC A120E mutant regardless of whether FRS2 was tyrosine-phosphorylated or not. Very low and equal amounts of FRS2 signal were detected in the all the lanes containing wild-type PKC or GST only after prolonged exposure of the blot (data not shown) and is unlikely to be significant. The association of PKC with FRS2 is therefore independent of tyrosine phosphorylation.
FRS2 Is Not an in Vitro Substrate of aPKCs-FRS2 has been reported to be serine/threonine-phosphorylated (4). We have also shown here that stimulation of Swiss 3T3 cells with FGF leads to the activation of PKC and the binding of FRS2 or SNT2 to regions in PKC and PKC that contain the catalytic domain. All the above evidence suggests that FRS2 is a likely substrate of PKC . We therefore addressed the enzyme-substrate relationship between PKC / and FRS2. We used hnRNPA1 and/or myelin basic protein (a general substrate for PKC) as positive controls. GST full-length FRS2 and/or FRS2 fragment Z were tested as substrates for the aPKCs. In vitro kinase assays, as described under "Experimental Procedures," were carried out using purified PKC or HA-tagged PKC fragment B (containing the kinase domain) or HA-tagged PKC A120E constitutively active mutant as a source of kinase activities. Fig. 6A (top left panel) revealed that hnRNPA1 and MBP are phosphorylated by purified PKC , but full-length FRS2 is not. In addition, although the HA-tagged PKC kinase domain and PKC A120E constitutively active mutant were active against MBP, they did not phosphorylate full-length FRS2 or FRS2 fragment Z, which is known to bind more strongly to PKC than the full-length FRS2 (Fig. 6A, lower  panels). To investigate whether post-translational modification of FRS2 was necessary for it to be a substrate of the aPKCs, FRS2 was immunoprecipitated from Swiss 3T3 cell lysates that had been either non-stimulated or stimulated with bFGF. In vitro kinase assays with PKC enzyme or with enzymatically active HA-tagged PKC proteins were carried out with the immunoprecipitated tyrosine-phosphorylated or non-phosphorylated FRS2. Again, the aPKC enzymes failed to phosphorylate either form of FRS2 (data not shown).
To show that the GST-FRS2 proteins could be an authentic substrate in vitro, GST full-length FRS2 was tested as a sub-strate for protein kinase A (PKA). FRS2 contains potential PKA phosphorylation sites (Ser-366 and Ser-429), and preliminary studies had shown that it was a likely substrate of that kinase. The hnRNPA1 protein was added as a positive control as it can also be phosphorylated by PKA (20). Myelin basic protein was also included in the assay. Fig. 6A (top right panel) shows that full-length GST-FRS2 could be phosphorylated by PKA to a greater degree than hnRNPA1. We have also performed in vitro kinase assays with purified PKC ␣ and PKC ␦ to determine whether FRS2 might be a substrate for PKCs from other subfamilies. Fig. 6B shows that whereas both PKC ␣ (left panel) and PKC ␦ (right panel) were able to phosphorylate MBP, they either failed to phosphorylate FRS2 (PKC ␣) or they phosphorylated FRS2 only very weakly (PKC ␦). Taken together, these results demonstrate that whereas FRS2 was phosphorylated by PKA, it is not an in vitro substrate and hence is an unlikely in vivo substrate for the PKCs. DISCUSSION The concept of non-receptor, tyrosine-phosphorylated proteins serving as dockers was conceived when the IRS family of proteins was discovered and characterized. Gradually, more proteins were identified that fall under this description. Like IRS proteins, Gab-1 and Dos possess modules such as pleckstrin homology or phosphotyrosine-binding (PTB) domains that help in their membrane localization (7). More recently, another docker protein named FRS2 has been identified and cloned (2). It has a myristoylation site for membrane association, an Nterminal phosphotyrosine-binding domain (PTB) for proteinprotein interactions, and multiple tyrosine residues that are targets for phosphorylation by the FGFRs upon ligand binding. FRS2 has been reported to bind to the SH2 domains of Grb2 (2) and SHP-2 in a tyrosine phosphorylation-dependent manner (8), and these associations exert a significant influence on the MAP kinase cascade. Docker proteins, including FRS2, have been shown to play important roles as initiation centers for diverse signaling pathways. Identification of proteins that bind to FRS2 will therefore contribute to the understanding of the signal transduction of FGF in cells.
In our studies, we found that about 5% of total FRS2 coimmunoprecipitated specifically with PKC , a member of a subfamily of the family of PKC kinases called the atypical PKCs (aPKCs). The interaction between FRS2 and PKC was shown to be mediated by a region (fragment B) in the aPKCs that encompasses the catalytic domain. We have also shown that activation of the aPKCs is necessary for its association with FRS2. It can be construed that the absence of co-immunoprecipitation of FRS2 with the cPKCs or nPKCs may be due to the inability of FGF to stimulate the various cPKCs and nPKCs. However, the B fragments of cPKC (e.g. PKC ␤II) and nPKC (e.g. PKC ␦) did not have the affinity to bind FRS2 and are unlikely to bind FRS2 even if they were activated. The sequence identities between the fragment B of PKC and those of other members were 86% for PKC and 44 -55% for other PKC members. The greater amino acid sequence homology between the aPKCs is sufficient to provide specificity for their binding to FRS2 but not for the more distantly related members of the PKC family.
The association between aPKCs and FRS2 is not that of an enzyme-substrate relationship. Only one strong in vivo substrate of the aPKCs has so far been identified. hnRNPA1 protein was identified in a yeast two-hybrid screen using the PKC kinase domain as bait (19). The optimal peptide sequence, determined by peptide library screening, for phosphorylation by aPKC is RRFKRQGS(P)FFYFF (where boldface indicates the motif required for phosphorylation and boldface italic indicates phosphorylation at serine) (23). This is similar to the motif on hnRNPA1 that surrounds the phosphorylated serine residue SQRGRSGS(P)GNFGG. It is crucial to have the basic and hydrophobic amino acid residues to the N and C terminus of the core sequence RXGS, respectively. Such a sequence was not found in FRS2, and this validates our experimental data. Although SNT2 seems to possess one such potential motif PL-TRRGS(P?)PRVFNFDF, it is only very weakly phosphorylated by the aPKCs (data not shown).
FRS2 is located at the plasma membrane of cells and associates with FGF receptors in a tyrosine phosphorylation-independent manner (5,6). It is possible that FRS2 may recruit the aPKCs to substrates in the vicinity of the FRS2-receptor complex at the cell-surface membrane. Proteins that are associated with FRS2 would therefore be potential substrates. SHP-2 has a potential PKC phosphorylation site, AGIGRTGT(P?)TFIVI (where P? indicates a potential threonine phosphorylation site). In vitro phosphorylation studies, however, showed that SHP-2 is not a substrate of PKC . Furthermore, yeast twohybrid assays did not show any association between PKC and SHP-2. The FGF receptor was excluded as a plausible substrate because it does not possess any potential aPKC phosphorylation sites in the cytosolic region. The identification and characterization of additional FRS2-associated proteins may lead to the identification of novel PKC substrates.
The association of aPKCs or SHP-2 with FRS2 has a common feature. Both the aPKCs and SHP-2 are enzymes but their binding partner, FRS2, is not a substrate for either the kinase or the phosphatase. On the other hand, the aPKCs and SHP-2 show differences in their manner of binding to FRS2. The binding of SHP-2, via its N-terminal SH2 domain, to FRS2 is dependent on tyrosine-phosphorylated residues in FRS2, and this interaction is necessary for the activation of the phosphatase (24). In contrast, the binding of PKC to FRS2 is not dependent on tyrosine phosphorylation of FRS2 but dependent on the activation of PKC . Hence, although FRS2 is both an activator and locator protein for SHP-2, it is likely to be only a locator protein for the activated aPKCs. However, we cannot rule out other possible functions FRS2 plays in the regulation of aPKCs such as post-translational modification.
Preliminary experiments suggested that FRS2 is unlikely to be an inhibitor of aPKC activity since incubation of a reaction mixture containing PKC and hnRNPA1 with FRS2 did not block phosphorylation of hnRNPA1 (data not shown). Although we cannot totally exclude the possibility that FRS2 is an inhibitor for aPKCs, we suspect that the role of FRS2 in binding the aPKCs is similar to that of a group of proteins called RACKs (Receptors for Activated C Kinase). The RACKs have been proposed to anchor PKC at specific locations in the cell (25,26). Like FRS2, these proteins have been shown to bind to the active conformation of the PKC but are themselves not substrates for the PKC.
All PKCs appear to be activated at the plasma membrane with phosphatidylserine being an important co-activator for all members of the greater family. In addition, the aPKCs are activated by particular inositol phospholipids (29). PKC has also received considerable recent attention as a target for PI-3 kinase (31,32). Two reports show that the phosphoinositidedependent protein kinase 1 (PDK-1), which binds with high affinity to the PI 3-kinase lipid product phosphatidylinositol 3,4,5-trisphosphate, phosphorylates and potently activates PKC along with two other substrates, also kinases, Akt/PKB and p70S6K. PKC and PDK-1 are associated in vivo, and membrane targeting of PKC renders it constitutively active in cells. The association between PKC and PDK-1 reveals extensive cross-talk between enzymes in the PI 3-kinase pathway (28,30). Evidence has been presented previously indicating that the IRS protein associates with enzymes involved in the PI 3-kinase pathway (7). It is possible that the strategic membrane locations of the IRS and FRS2 docker proteins may see them playing a central role in the PI 3-kinase pathway as well as the MAP kinase pathways and the likely interactions between the two pathways.
In conclusion, the association of activated aPKCs to FRS2 may serve to target the aPKCs to specific sites on the plasma membrane where their substrates are located or where their activities are regulated. Interestingly, the three proteins Grb2, SHP-2, and PKC / that are recruited by FRS2 to the membrane are implicated in the MAP kinase pathway at different positions (either upstream or downstream of Ras). The association may be part of a large multimeric signaling complex where signals can be integrated and "fine-tuned," resulting in their propagation from the cell surface to the nucleus.