Shc and CEACAM1 Interact to Regulate the Mitogenic Action of Insulin*

CEACAM1, a tumor suppressor (previously known as pp120), is a plasma membrane protein that undergoes phosphorylation on Tyr 488 in its cytoplasmic tail by the insulin receptor tyrosine kinase. Co-expression of CEACAM1 with insulin receptors decreased cell growth in response to insulin. Co-immunoprecipitation experiments in intact NIH 3T3 cells and glutathione S -trans-ferase pull-down assays revealed that phosphorylated Tyr 488 in CEACAM1 binds to the SH2 domain of Shc, another substrate of the insulin receptor. Overexpressing Shc SH2 domain relieved endogenous Shc from binding to CEACAM1 and restored MAP kinase activity, growth of cells in response to insulin, and their colonization in soft agar. Thus, by binding to Shc, CEACAM1 sequesters this major coupler of Grb2 to the insulin receptor and down-regulates the Ras/MAP kinase mitogenesis pathway. Additionally, CEACAM1 binding to Shc enhances its ability to compete with IRS-1 for phosphorylation by the insulin receptor. This leads to a decrease in IRS-1 binding to phosphoinositide 3 (cid:1) -kinase and to the down-regulation of the phosphoinositide 3 (cid:1) -kinase/Akt

Insulin binding to its receptor activates its tyrosine kinase to cause phosphorylation of the receptor and of endogenous substrates, including CEACAM1 (previously known as pp120) (1), insulin receptor substrate proteins (IRS-1-4), 1 Shc (2, 3), and others. Phosphorylation of different substrates is required to mediate the diverse effects of hormones on metabolism and growth (4 -6).
Shc is a Src homology 2 (SH2)-containing cytoplasmic adaptor protein that undergoes phosphorylation by receptors of the tyrosine kinase family (7). Activation of receptors causes redistribution of Shc from the perinuclear region to the cytosolic site of the plasma membrane (8). The tyrosine kinase receptors that phosphorylate Shc include insulin and insulin-like growth factor 1 (IGF-1) receptors (2,9). The Shc family of proteins consists of three isoforms. p46/p52 that are ubiquitously expressed are the products of the same transcript and result from alternative usage of two in-frame ATGs. In contrast, p66 that is mostly expressed in epithelial cells is translated from a different transcript (10,11). The three isoforms have overlapping domains as follows: an SH2 domain at the C terminus, an adjacent glycine/proline-rich collagen homology (CH1) domain, and a phosphotyrosine binding (PTB) domain in the N terminus of p46/p52. p66 Shc contains an additional collagen homology (CH2) domain at its N terminus end.
By binding to other signaling proteins, Shc exerts many effects on the cell. Upon its phosphorylation on Tyr 317 in the CH1 domain by the insulin receptor, Shc binds to the SH2 domain of Grb2 and couples it to the receptor (12,13). This leads to the association of Grb2 with the Son of Sevenless Ras GDP/GTP exchanger, causing translocation of Son of Sevenless to the plasma membrane in proximity to its p21 ras substrate (14), activation of the Ras/mitogen-activated protein kinase (MAP kinase) pathway, and regulation of cell growth, differentiation, and proliferation in response to insulin and other growth factors. Shc also interacts with adaptins to play a role in growth factor endocytosis (15). It also binds to cadherins to regulate the role of these proteins in cell-cell adhesion and cell morphogenesis (16,17) and to integrins, although indirectly, to regulate cell proliferation and cell cycle progression (18).
Members of the IRS family do not contain SH2 domains but transmit insulin signaling by forming complexes via their multiple phosphotyrosine-containing binding motifs with SH2 domains in signaling molecules such as the growth factor receptor-binding protein (Grb2) (19,20), Syp (SH PTP2) phosphotyrosine phosphatase (21), and phosphatidylinositol (PI)-3Ј kinase (22). Coupling of PI-3 kinase to the insulin receptor by the IRS proteins activates downstream signaling molecules like Akt and p70 ribosomal protein 6 kinase (p70 S6 kinase). Akt activation promotes cell growth and proliferation (23,24) in addition to mediating anti-apoptosis and cell survival (25,26). Activation of p70 S6 kinase mediates the mitogenic effects of insulin in many cell types, including hepatocytes (27).
Among insulin-targeted tissues, CEACAM1 is only expressed in liver. It is a plasma membrane glycoprotein of M r ϳ120,000. The rat protein is expressed as two spliced variants differing by the inclusion (CEACAM1-4L) or exclusion (CEACAM1-4S) of a 61-amino acid segment in the C terminus of its cytoplasmic domain (28). The truncated isoform lacks all phosphorylation sites. Site-directed mutagenesis in NIH 3T3 cells revealed that CEACAM1 is constitutively phosphorylated on Ser 503 and that this phosphorylation is required for its phosphorylation on the Tyr 488 residue by the insulin receptor tyrosine kinase (1).
The function of CEACAM1 remains elusive. It may function as a tumor suppressor in breast, colon, bladder, liver, and prostate (29 -32) and as a down-regulator of the mitogenic action of insulin (33,34). CEACAM1 may up-regulate the transport of bile acids and insulin (33,35) in the hepatocyte. Supportive evidence for a role of CEACAM1 in cell adhesion has also emerged (36,37).
The basic mechanism of CEACAM1 functions is not completely understood. However, CEACAM1 phosphorylation is required for its function in bile acid and insulin transport (38,39) and in tumor suppression (40). Failure of phosphorylationdefective CEACAM1 mutants to decrease cell growth in response to insulin suggested that CEACAM1 phosphorylation is required for its down-regulatory effect on insulin mitogenesis (33). This was further supported by the observation that CEACAM1 failed to down-regulate growth of cells overexpressing IGF-1 receptors that do not phosphorylate CEACAM1 (34,41).
Because sequences flanking Tyr 488 of CEACAM1 (Tyr-Ser-Val-Leu) closely resemble binding sites of Shc SH2 domains from several proteins with important functions in the immune system, human CD3 ␥ and ␦ chain (Tyr-Ser-His-Leu), the human CD3 chain (Tyr-Asp-Val-Leu), the mouse CD3 ⑀ chain (Tyr-Ser-Gly-Leu), and the mouse Ig ⑀ receptor b (Tyr-Ser-Glu-Leu) (42), it is possible that phosphorylated CEACAM1 binds to Shc. Co-immunoprecipitation and GST pull-down assays revealed that insulin stimulates binding of phosphorylated Tyr 488 in CEACAM1 to the SH2 domain of Shc. Overexpressing the Shc SH2 domain in NIH 3T3 cells transfected with CEACAM1 released endogenous Shc proteins from the CEACAM1 complex and restored cell growth and MAP kinase activity in response to insulin. This suggests that by binding to Shc, CEACAM1 sequesters it and limits Grb2 coupling to the insulin receptor. Additionally, this binding increases the ability of Shc to compete with IRS-1 for phosphorylation by the insulin receptor in light of the fact that this event requires an intact Tyr 960 in the juxtamembrane domain of the receptor (43,44), as opposed to that of CEACAM1 which requires an intact Tyr 1316 in the C terminus of the ␤-subunit of the receptor (34). Decreased IRS-1 phosphorylation leads to down-regulation of the PI-3 kinase/Akt mitogenesis pathway in cells co-expressing CEACAM1. Thus, it appears that Shc binding underlies the down-regulatory effect of CEACAM1 on cell growth and proliferation.

EXPERIMENTAL PROCEDURES
Materials-Reagents for polyacrylamide gel electrophoresis were purchased from Bio-Rad, and those for immunoblotting and for the glutathione S-transferase (GST) system were from Amersham Biosciences. The baculovirus-purified ␤-insulin receptor kinase (aa 941-1343) of the cytoplasmic tail of ␤-subunit of the insulin receptor was from Calbiochem. Polyclonal antibodies against the insulin receptor ␤-subunit (polyclonal) and Grb2 were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), and those against IRS-1 (polyclonal), the Nterminal SH2 of the p85 subunit of PI-3 kinase (monoclonal), Shc (polyclonal), and phosphotyrosines (monoclonal) were from Upstate Biotechnology, Inc. (Lake Placid, NY). The monoclonal antibody used to immunoprecipitate CEACAM1 (␣-HA4, an identical protein to CEACAM1) was purified from ascites fluid from HA4 c19 cells purchased from the Developmental Studies Hybridoma Bank (Department of Biology, University of Iowa, Iowa City, IA). ␣-76 Ex polyclonal antibody was raised in rabbit against aa 51-64 in the extracellular domain of CEACAM1. The polyclonal antibody used in affinity purification of CEACAM1 fusion peptide (␣-Tyr-513) was raised in rabbit against peptide (aa 505-520) in the intracellular domain of rat liver CEACAM1.
Construction of Expression Vectors-Amplification and subcloning of the cDNA molecules encoding wild type (WT) CEACAM1-4L and -4S (referred to as -L and -S for simplicity) and site-directed CEACAM1-L mutants (Y488F, Y488F/Y513F, Y513F, and S503A) into a bovine papilloma virus-based expression vector (Amersham Biosciences) at the XhoI/NotI sites were described previously (1). Subcloning of the Shc SH2 domain (aa 367-473) into the cytomegalovirus-based pcDNA3.1/V5-His B expression vector (Invitrogen, Carlsbad, CA) was achieved by excising its encoding DNA from the pGEX-2T construct (45) by EcoRI-BamHI digestion and in-frame ligating it into the expression plasmid at the same sites. As described previously (28), the cDNA fragment (nucleotides 258 -554) encoding the SH2 domain of Grb2 (aa 58 -152) was amplified in a polymerase chain reaction using the recombinant Grb2 cDNA (in the pbluescript SK vector from ATCC, Manassas, VA) as template, and oligonucleotides S-258 (gagtcgggatcc-TGGTTTTTTGGCAAAATCCCC) and ␣-554 (gagtcggaattcTGGCTGCT-GTGGCATCTGTTC) as sense and antisense primers, respectively. The sense oligonucleotide contained a BamHI and the antisense primer contained an EcoRI restriction site (shown in lowercase letters) to allow for in-frame subcloning of the cDNA product into the pcDNA3.1/V5-His B expression vector.
Expression of Peptides in the GST Fusion Protein System-Synthesis and amplification of GST fusion peptides of the intracellular domain of CEACAM1 and Shc peptides were described previously (1,15). Competent Escherichia coli HB101 cells (Invitrogen) were transformed with the GST fusion vectors, and the resulting GST peptides were coupled to 50% reduced glutathione-Sepharose 4B, as described previously (1).
Phosphorylation of the GST-CEACAM1 Fusion Peptide by the Baculovirus-purified Insulin Receptor Tyrosine Kinase-The GST-CEACAM1 fusion peptide was treated with thrombin (Amersham Biosciences) to release the GST moiety followed by affinity purification off CEACAM1 antibody-coupled agarose. As described previously (1), 0.1 g of CEACAM1 intracellular peptide was phosphorylated by the ␤-insulin receptor kinase (aa 941-1343) (10 units) in the presence of [␥Ϫ 32 P]ATP (6000 Ci/mmol; PerkinElmer Life Sciences). 10 g of Sepharose-coupled GST-Shc were then added and mixed at 4°C for 30 min. The Sepharose pellet was washed in HNTG buffer (150 mM Hepes, pH 7.6, 50 mM NaCl, 0.5% Triton X-100, 10% glycerol) and analyzed by 10% SDS-PAGE and autoradiography.
Cell Culture and Transfections-Stable transfection of NIH 3T3 cells with bovine papilloma virus-based expression vector-human insulin receptors (hIR) (with or without wild type and mutant isoforms of CEACAM1) in the presence of the neomycin-resistant (Neo r ) gene was described previously (1,39). Stable co-expression of the SH2 domains of Shc and Grb2 in NIH-3T3 cells expressing ϳ1.0 -1.2 ϫ 10 5 hIR per cell with or without CEACAM1-L in the presence of pREP4-hygromycinresistant (Hygro r ) was achieved by the LipofectAMINE method, as described previously (1). Isolated clones were expanded, maintained in medium containing hygromycin (200 g/ml) (Invitrogen), and lysed in 1% Triton X-100 prior to protein analysis by 12% SDS-PAGE, as described previously (1). Screening for SH2-Shc or SH2-Grb2 expression was attained by immunoblotting with polypeptide antibodies against the V5 epitope and His (Invitrogen).
Primary Hepatocyte Cultures-6-Month-old male mice were anesthetized with sodium pentobarbital (30 g/g body weight). Hepatocytes were isolated by a two-stage collagenase perfusion through the portal vein (1 mg/ml collagenase (Roche Molecular Biochemicals) in Leibovitz's L-15 media (Invitrogen) supplemented with glucose (1 mg/ml)). The viability of hepatocytes was determined by trypan blue dye exclusion. Cells were then plated at 4.5 ϫ 10 6 cells per 100-mm dish in 10 ml of Dulbecco's modified Eagle's medium, supplemented with fetal bovine serum (10% v/v), and penicillin/streptomycin (1% v/v) at 37°C. Cultures were refed with 10 ml of medium after a 2-h attachment period (46).
Co-immunoprecipitation in Intact Cells-Following overnight starvation of serum, transfected cells were treated with insulin (100 nM) for 8 min prior to lysis, as described previously (1). Proteins were immunoprecipitated with antibodies against CEACAM1, Shc, IRS-1, or Grb2, and the immunopellets were washed with phosphate-buffered saline, pH 7.4, prior to electrophoresis through SDS-PAGE and immunoblotting with the same antibody used in immunoprecipitation (to account for the amount of antigen in the immunopellet) and with the antibody against the other protein that may have co-immunoprecipitated. Proteins were re-immunoprobed with ␣-actin antibody to control for the amount of proteins. Following horseradish peroxidase labeling, proteins were detected by enhanced chemiluminescence.
Quantitative Immunoprecipitation-Primary hepatocytes were insulin-treated prior to lysis. Proteins were subjected to three sequential immunoprecipitations with an antibody against BGP1, the mouse homolog of CEACAM1, to immunodeplete CEACAM1. 300 g of proteins were trichloroacetic acid-precipitated from the final supernatant and subjected to analysis by 6 -12% gradient SDS-PAGE. Proteins were then sequentially immunoprobed with antibodies against Shc and CEACAM1. Equal amounts of proteins were trichloroacetic acid-precipitated from cell lysates and analyzed by the same SDS-PAGE as measured by the total pool of Shc and CEACAM1 proteins in primary hepatocytes.
Cell Growth and Proliferation-As described previously (34), cells (3 ϫ 10 3 cells per well) were rendered quiescent in the absence of serum, incubated with insulin (10 nM) or complete medium for 24 and 48 h, trypsinized, and counted in a Coulter Counter (Z1 model). Hormoneinduced cell growth was calculated as percent maximal minus basal growth divided by the number of cells grown in complete medium.
MAP Kinase Assay-4 ϫ 10 6 cells were insulin-treated for 0 -90 min and assayed for MAP kinase activity per manufacturer's instructions (Cell Signaling Technology, Inc; Beverly, MA). Following analysis by 10% SDS-PAGE, proteins were immunoblotted with the phospho-MAP kinase antibody and detected by the LumiGlo detection system (New England Biolabs, Beverly, MA). To account for the amount of Erk1 (p44 MAP kinase) and Erk2 (p42 MAP kinase) in the immunopellet, proteins were reimmunoblotted with a p44/p42 MAP kinase antibody.
Akt Phosphorylation Assay-Cells were incubated with 0.5% fetal bovine serum, insulin-treated for 0 -60 min, lysed in SDS buffer, and analyzed by 10% SDS-PAGE per the manufacturer's instructions (Cell Signaling Technology). Akt was detected by immunoblotting with antiphospho-Ser 473 Akt antibody and the LumiGlo detection system.
PI-3 Kinase Assay-300 -450 g of homogenates derived from insulin-treated cells were immunoprecipitated with anti-phosphotyrosine antibodies (PY20) (Santa Cruz Biotechnology, Inc.). The immunopellets were washed and resuspended in reaction buffer (25 mM Hepes, pH 7.1, 0.5 mM EGTA, and 0.5 mM sodium phosphate) with 10 mg of phosphatidylinositol. The phosphorylation reaction was initiated by 10 l of 250 M ATP containing 5 Ci of [␥-32 P]ATP and incubated for 6 min at room temperature. The reaction was stopped by adding 15 l of 4 N HCl. Phospholipids were then extracted with 130 l of CHCl 3 /methanol (1:1), and 30 l of the CHCl 3 layer was resolved on thin layer chromatography plates (47).
Quantitation of Proteins-Autoradiograms were scanned on an imaging densitometer (Bio-Rad model GS-670), and the proteins were quantitated with the Image NIH version 1.61 Macintosh software program.
Statistical Analysis-Curves were compared by a multivariate analysis of variance, and individual points were compared by paired t tests. p values of less than 0.05 were considered statistically significant.

Association between CEACAM1 and Shc in Intact Cells-Co-
immunoprecipitation assays were employed to examine whether CEACAM1 and Shc form a complex in vivo (Fig. 1). Untransfected NIH 3T3 cells and cells stably expressing wild type insulin receptors alone (WT IR) or co-expressing either full-length (WT IR/CEACAM1-L) or truncated CEACAM1 (WT IR/CEACAM1-S) were insulin-treated prior to cell lysis. Immunoprecipitation was carried out with ␣-Shc ( Fig. 1, lanes 1-10), ␣-CEACAM1 antibodies (␣-CC1) (Fig. 1, lanes 11 and 12), and with normal rabbit globulin (data not shown). Proteins were immunoblotted with ␣-Shc (Fig. 1, panel ii) and reprobed with ␣-CC1 (Fig. 1, panel i) to account for the amount of Shc and CEACAM1 in the immunopellets, respectively. Additionally, gels were reprobed with ␣-actin antibody to control for the amount of proteins in the immunopellets (panel iii). The nonspecific band of ϳ75 kDa was called that because it was detected in normal rabbit globulin immunopellets (not shown). As Fig. 1 reveals, CEACAM1 was detected in Shc, but not in normal rabbit globulin immunopellets, from untreated WT IR/ CEACAM1-L stable transfectants (with a and b being two different stable clones) (Fig. 1, panel i, lanes 5 and 7). Normalization for the amount of Shc in the immunopellets (Fig. 1, panel ii) revealed that insulin treatment yielded a 2-3-fold increase in the amount of CEACAM1-L co-immunoprecipitated with Shc in these transfectants (Fig. 1, panel i, lanes 6 versus 5 and 8 versus 7). Identical results were obtained when the ␣-Shc antibody was used to immunoblot CEACAM1 immunoprecipitates from cells co-expressing CEACAM1-L ( Fig. 1, panel ii, lane 12 versus 11). In contrast, CEACAM1-S was not detected in Shc immunopellets (Fig. 1, panel i, lanes 9 and 10). Failure to detect CEACAM1 in either untransfected cells (Fig.  1, panel i, lanes 1 and 2) or in cells transfected with insulin receptors alone (WT IR) (Fig. 1, panel i, lanes 3 and 4) suggests that the association between CEACAM1-L and Shc is specific. This suggests that CEACAM1 constitutively associates with Shc in vivo via its intracellular domain and that this association is increased with insulin-stimulated phosphorylation of CEACAM1 by the receptor.
Identification of Shc-binding Sites in CEACAM1-L-Because CEACAM1 is basally phosphorylated on Ser 503 (1), we abolished this phosphorylation site by mutating Ser 503 to alanine, and we examined the effect of this mutation on CEACAM1/Shc interaction. In the absence of insulin, p46 Shc was detected in ␣-CC1 immunopellets derived from cells expressing WT CEACAM1-L with ( Fig. 2A, panel i, lane 3) or without insulin receptors ( Fig. 2A, panel i, lane 1). In contrast, p46 Shc coimmunoprecipitation with CEACAM1 was markedly reduced when Ser 503 in CEACAM1-L was mutated to alanine (S503A CEACAM1-L) ( Fig. 2A, panel i, lane 5 versus 3 or 1). This suggests that CEACAM1 constitutively associates with p46 Shc and that its basal phosphorylation on Ser 503 regulates this association. Insulin treatment increased the amount of p46 Shc in the CEACAM1 immunopellet by 4 -5-fold in cells co-expressing WT ( Fig. 2A, panel i, lane 4 versus 3) but not S503A CEACAM1-L ( Fig. 2A, panel i, lane 6 versus 5). Similar results were obtained when ␣-Shc immunopellets were immunoblotted with ␣-CEACAM1 antibody (data not shown). Because phosphorylation of CEACAM1 on Ser 503 is required for its phosphorylation on Tyr 488 by the insulin receptor kinase, these data suggest that CEACAM1 binding to Shc is up-regulated by its tyrosine phosphorylation. Thus, we examined the effect of abolishing CEACAM1 phosphorylation on its interaction with Shc. To this end, we mutated Tyr 488 to nonphosphorylatable phenylalanine (Y488F). Contrary to cells expressing the phosphorylatable isoforms of CEACAM1-L (WT and the Y513F mutant) (Fig. 2B, panel i, lanes 2 versus 1 and 6 versus 5, respectively), insulin did not increase the amount of p46 Shc in the CEACAM1 immunopellets from cells co-expressing CEACAM1 phosphorylation-defective isoforms (Y488F and Y488F/Y513F) (Fig. 2B,  panel i, lanes 8 versus 7 and 4 versus 3, respectively). Thus, it appears that CEACAM1 phosphorylation increases its binding to Shc on Tyr 488 . Ser 503 (and to some extent Tyr 513 ) could bestow on the intracellular domain of CEACAM1 the proper conformation for its interaction with Shc at the basal state.
Identification of CEACAM1-binding Sites in Shc-Because Shc phosphorylation requires an intact Tyr 960 in the juxtamembrane domain of the insulin receptor, we examined the effect of mutating Tyr 960 to phenylalanine on Shc association with CEACAM1. Insulin increased the level of CEACAM1 in the Shc immunopellets by ϳ7-8-fold in cells co-expressing WT or Y960F insulin receptors (Fig. 3A, panel i, lanes 2 versus 1  and 4 versus 3, respectively), as corrected for the amount of Shc immunoprecipitated (Fig. 3A, panel ii, lanes 1-4). Similarly, insulin induced by 2-3-fold the amount of Shc (in particular p46) co-precipitated with CEACAM1 immunopellets in cells co-expressing WT or Y960F receptors (Fig. 3A, panel ii, lanes 6  versus 5 and 8 versus 7, respectively). This suggests that Shc phosphorylation is not required for its binding to CEACAM1.
To identify the CEACAM1-binding sites in Shc, we employed the GST pull-down assay in which a peptide encompassing the intracellular domain of CEACAM1 (ϳ7 kDa) was prephosphorylated by the insulin receptor kinase (␤IR) (Fig. 3B, evennumbered lanes) or buffer alone (Fig. 3B, odd-numbered lanes) in the presence of [␥Ϫ 32 P]ATP prior to its incubation with Sepharose-coupled GST-Shc fusion peptides. These peptides include full-length Shc (aa 4 -473, lanes 3 and 4) or deletion  11 and 12) were removed. The CEACAM1 peptide appeared to be phosphorylated (and bound to GST-Shc fusion peptides) even in the absence of ␤IR (Fig. 3B, panel i,  lanes 3, 7, 9, and 11), suggesting co-purification of a bacterially derived kinase that is capable of phosphorylating CEACAM1, but not the GST portion (Fig. 3B, panel i, lane 1), in this in vitro system. Regardless, the absence of CEACAM1 in the GST precipitate lacking Shc (Fig. 3B, panel i, lanes 1 and 2) suggests that CEACAM1 specifically bound to the Shc moiety of the GST-Shc fusion peptides (lanes 3-12). Phosphorylation of CEACAM1 by the insulin receptor kinase induced a 2-3-fold increase in its co-precipitation with the full-length GST-Shc fusion peptide (Fig. 3B, panel i, lane 4 versus 3) or with peptides from which the CH1 or the PTB domains were deleted either individually (Fig. 3B, panel i, lanes 8 versus 7 and 10 versus 9) or collectively (Fig. 3B, panel i, lanes 12 versus 11). In contrast, when the SH2 domain was deleted, CEACAM1 did not co-precipitate with the GST-Shc fusion peptide (Fig. 3B,  panel i, lanes 5 and 6). This suggests that the SH2 domain of Shc is the main site of its direct interaction with CEACAM1. That CEACAM1 intracellular peptide bound Shc even in the absence of Shc phosphorylation, which chiefly occurs in the CH1 domain (Fig. 3B, panel i, lanes 7 and 8 and 11 and 12), strengthens the conclusion that Shc binding to CEACAM1 occurs independently of Shc phosphorylation. Notably, ␤IR coprecipitated with full-length Shc fusion peptide (Fig. 3B, panel  i, lane 4). This suggests that Shc, IR, and CEACAM1 form a complex, with Shc mediating the indirect interaction between CEACAM1 and IR. Interestingly, ␤IR was not detected in pellets of GST peptides containing PTB but lacking SH2 and CH1 domains of Shc (Fig. 3B, lanes 5-8). This suggests that these two domains confer stability or proper conformation for Shc binding to the receptor at its PTB domain.
Shc Associates with CEACAM1 in Hepatocytes-To examine whether CEACAM1 binds to Shc in liver, the site of CEACAM1 expression, we immunodepleted CEACAM1 in cell lysates of mouse primary hepatocytes prior to immunoprecipitating with ␣-Shc antibody to measure the amount of Shc escaping binding to CEACAM1 (Fig. 4, lanes 1 and 2). Equal amounts of proteins were trichloroacetic acid-precipitated from undepleted cell lysates to measure total Shc and CEACAM1 pools (Fig. 4, lane 3). Following analysis by SDS-PAGE, proteins were immunoblotted with antibodies against Shc or CEACAM1. As Fig. 4 reveals, depleted CEACAM1 appeared to co-precipitate ϳ6% of p46 Shc in the absence of insulin and ϳ68% in its presence, as indicated by the 94 (Fig. 4, lane 1 versus 3) and 32% (Fig. 4, lane  2 versus 3) of the total p46 Shc pool remaining in the supernatant. In contrast to p46 Shc , ϳ60% of p66 Shc and p52 Shc bound to CEACAM1 in the absence (Fig. 4, lane 1 versus 3) and presence (Fig. 4, lane 2 versus 3) of insulin. This suggests that tyrosine phosphorylation of CEACAM1 preferentially increased its binding to p46 Shc in primary hepatocytes.
Regulation of Cell Proliferation by CEACAM1-As expected from our previous studies (34), expressing CEACAM1-L decreased cell growth in response to insulin by ϳ7-fold as compared with cells expressing insulin receptors alone (Fig. 5A, panel i, P IR/CC1 , 1.5 Ϯ 1.0 versus P IR , 14.7 Ϯ 2.1; p Ͻ 0.05). The decrease in cell growth was associated with an increase in the amount of p46 Shc co-precipitated with CEACAM1 (Fig. 5A, panel iii, ϩ versus Ϫlane in P IR/CC11 ). We hypothesized that CEACAM1 binds to Shc and sequesters this coupler of Grb2 to the receptor, thus down-regulating the MAP kinase mitogenesis pathway and reducing cell proliferation in response to insulin. To test this hypothesis, we investigated the effect of decreasing Shc binding to CEACAM1, by overexpressing its SH2 domain on insulin-stimulated cell growth. Overexpressing the Shc SH2 domain relieved p46 Shc from binding to CEACAM1 in response to insulin in cells co-expressing insulin receptors and CEACAM1 (Fig. 5A, panel iii, ϩ lanes in clones (4 -6) by comparison to the ϩ lane in P IR/CC1 parent cells). Furthermore, it restored the growth of these clonal cells in response to insulin to the same level observed in cells expressing insulin receptors alone (Fig. 5A, panel i, 17.0 Ϯ 1.6, 18.0 Ϯ 3.2, 15.9 Ϯ 1.7 in clones 4 -6, respectively versus 14.7 Ϯ 2.1 in P IR ; p Ͼ 0.05). The restorative effect of the Shc SH2 domain was specific to cells co-expressing CEACAM1 insofar as it failed to increase insulin-induced growth in cells expressing insulin receptors alone (Fig. 5A, panel  parent clone; p Ͼ 0.05). Similar results were obtained when 0.1 nM insulin was used (not shown). To assess further the specificity of the "competitive inhibition" of Shc binding to CEACAM1 by the overexpressed Shc SH2 domain, we transfected cells with the SH2 domain of Grb2, and we examined its effect on cell growth in response to 10 nM insulin. As expected, overexpressing the Grb2 SH2 domain markedly decreased Grb2 co-immunoprecipitation with Shc (Fig. 5B, panel iii, ϩ lane versus Ϫlane in clones 5 and 13 by comparison to their corresponding P IR/C and P IR parent cells, respectively). This was associated with inhibition of the stimulatory effect of insulin on cell growth in all cells tested (Fig. 5B, panel i).
Next, we investigated whether overexpressing the Shc SH2 domain alters CEACAM1 regulation of soft agar colonization of insulin receptor-transfected NIH 3T3. As shown in Fig. 5C, cells co-transfected stably with CEACAM1-L (P IR/CC1 ) exhibited 50% of the transformation capacity of those devoid of CEACAM1-L expression (P IR ) ( 4. Shc/CEACAM1 association in hepatocytes. Mouse primary hepatocytes were allowed to grow for 24 h in complete medium. Following overnight starvation of serum, cells were treated with insulin (ϩ) or buffer alone (Ϫ) for 8 min, lysed, and subjected to three sequential immunoprecipitation with an antibody against the mouse homolog of CEACAM1 (␣-mCC1) (lanes 1 and 2) to immunodeplete CEACAM1. Equal amounts of proteins were trichloroacetic acid-precipitated from the final supernatant and from cell lysates and subjected to analysis by 6 -12% gradient SDS-PAGE followed by immunoblotting with ␣-Shc and ␣-CC1 antibody. For photographic clarity, only CEACAM1 and Shc bands are shown. This represents at least three experiments. P IR ; p Ͼ 0.05). Furthermore, this effect of the Shc SH2 domain was specific to cells co-expressing CEACAM1 as it failed to alter transformation of cells expressing insulin receptors alone (21.67 Ϯ 2.08, 26.67 Ϯ 2.53, 25.33 Ϯ 3.05 in clones 44, 45, and 48, respectively, versus 21.33 Ϯ 5.51 in P IR ; p Ͼ 0.05). Thus, it appears that Shc binding mediates, at least partially, the regulation of cell tumorigenicity by CEACAM1.
Down-regulation of the MAP Kinase Mitogenesis Pathway in the Presence of CEACAM1-We then tested whether by binding to Shc, CEACAM1 sequesters it to down-regulate MAP kinase activity and cell proliferation in response to insulin. To this end, we treated quiescent cells for 0 -90 min with insulin (100 nM) and measured MAP kinase activity by immunoblotting proteins with an antibody against phospho-MAP kinase 44 (ERK1) and 42 (ERK2) (Fig. 6A, panel i). MAP kinase activity was normalized per the amount of the enzyme in the immunopellet (Fig. 6A, panel ii) and summarized in Fig. 6A, panel  iii). Insulin-stimulated ERK1 activity was more intense and prolonged in cells expressing insulin receptors alone (IR) than in cells co-expressing CEACAM1-L (IR/CC1) (Fig. 6A, panel iii,  open squares versus open circles). Although expressing the SH2 domain of Shc did not significantly alter ERK1 activity in cells expressing insulin receptors alone (Fig. 6A, panel iii, solid versus open squares), it increased the level of ERK1 activation and prolonged it over the entire period examined in cells coexpressing CEACAM1 (Fig. 6A, panel iii, solid versus open  circles). Similar results were obtained for ERK2 and other clones (not shown). This supports the hypothesis that CEACAM1 binding to Shc sequesters it and limits the activation of the Ras/MAP kinase mitogenesis pathway in response to insulin.
Because IRS-1 also couples Grb2 to the insulin receptor, albeit to a lesser extent than Shc (20), we examined the effect of CEACAM1 binding to Shc on IRS-1/Grb2 interaction. As indicated in Fig. 6B, Grb2 binding to both Shc (panel i) and IRS-1 (panel iv) in response to insulin was decreased in cells co-expressing CEACAM1 (P IR/CC1 ) as opposed to cells expressing insulin receptors alone (P IR ) (ϩ versus Ϫlane). Furthermore, relieving Shc binding to CEACAM1 by overexpressing the Shc SH2 domain restored insulin-stimulated Shc and IRS-1 binding to Grb2 in cells co-expressing CEACAM1 (panels i and iv, ϩ versus Ϫlane in clone 4 as compared with ϩ versus Ϫ in parent P IR/CC1 cells). Thus, it appears that not only did CEACAM1 binding to Shc reduce its interaction with Grb2, but it also adversely affected the interaction of IRS-1 with Grb2. This could partially mediate the decrease in MAP kinase activity in the presence of CEACAM1 (Fig. 5A, panel i).
Because IRS-1/Grb2 interaction is mediated by IRS-1 phosphorylation, we examined insulin-stimulated IRS-1 phosphorylation in the presence of CEACAM1. Increased incorporation of phosphorylated tyrosines by insulin treatment for 2-20 min over the basal state (at time 0) was calculated and plotted (Fig.  6C, panels iii and iv). Interestingly, CEACAM1 co-expression shortened the duration of IRS-1 phosphorylation (panels ii and iv) as it slightly increased the extent of Shc phosphorylation (panels i and iii).
Down-regulation of the PI-3 Kinase/Akt Pathway in the Presence of CEACAM1-Next, we examined whether the decrease in IRS-1 phosphorylation adversely affected PI-3 kinase activation. Correcting for the amount of IRS-1 (Fig. 7A, panel ii) revealed that insulin treatment significantly lowered the level of PI-3 kinase in the IRS-1 immunopellet in cells co-expressing CEACAM1 by comparison to cells expressing insulin receptors alone (Fig. 7A, panel i). This was associated with a 3-4-fold decrease in PI-3 kinase activity in cells co-transfected with CEACAM1 by comparison to cells transfected with insulin re-FIG. 5. Functional correlation. A, cell proliferation in response to insulin. NIH 3T3 cells were transfected with WT IR either alone (P IR ) or in addition to CEACAM1-L (P IR/CC1 ). The SH2 domain of Shc was stably transfected in each cell line, and three different resulting clones were examined (clones 44, 45, and 48 from cells expressing IR and SH2-Shc, and clones 4 -6 from cells expressing IR, CEACAM1-L. and SH2-Shc) (panel ii). Following incubation for 24 h in serum-containing complete medium (to determine maximum growth) or in serum-free medium supplemented with 0.1% bovine serum albumin either alone (to determine basal growth) or with 10 nM insulin, cells were trypsinized and counted (panel i). These experiments were performed in triplicate and repeated at least three times. Data represent the mean ؎ S.D. of these repeated experiments. Asterisk denotes p Ͻ 0.05 in P IR/CC1 versus P IR parent clones. Panel iii, co-immunoprecipitation of CEACAM1 and Shc was examined in cells co-expressing insulin receptors and CEACAM1 without (P IR/CC1 ) and with the Shc SH2 domain (clones 4 -6). B, NIH 3T3 cells were transfected with WT IR either alone (P IR ) or in addition to CEACAM1-L (P IR/CC1 ). The SH2 domain of Grb2 was stably transfected in each of these cell lines, and two different resulting clones were examined (clones 13 and 14 from cells expressing IR and SH2-Shc, and clones 5, and 9 from cells expressing IR, CEACAM1-L, and SH2-Shc) (panel ii). Cell counts in response to 10 nM insulin were measured as in A (panel i). These experiments were performed in triplicate and repeated at least three times. Data represent the mean ؎ S.D. of these repeated experiments. Asterisks denote p Ͻ 0.05, with one in P IR/CC1 versus P IR parent cells, and two in SH2 Grb2 clones versus parent cells. Panel iii, co-immunoprecipitation of Grb2 and Shc was examined. C, soft agar colony formation was assayed in NIH 3T3 cells transfected with WT IR alone (P IR ) or in addition to SH2-Shc (clones 44, 45, and 48), to CEACAM1-L (P IR/CC1 ), and to CEACAM1 with SH2-Shc (clones 4 -6). These experiments were performed in triplicate and repeated at least three times. Data represent the mean ؎ S.D. of these repeated experiments. Asterisk denotes p Ͻ 0.05 in P IR/CC1 versus P IR parent cells.
ceptors alone (Fig. 7A, panels iv and v). Furthermore, the decrease in PI-3 kinase activity in cells co-expressing CEACAM1 was accompanied by a decrease in Akt phosphorylation as compared with cells expressing insulin receptors alone (Fig. 7B, panel i). DISCUSSION CEACAM1, a substrate of the insulin receptor kinase, decreases the mitogenic action of insulin. This effect requires its phosphorylation by the receptor (33,34). The current studies reveal that CEACAM1 phosphorylation increases its association with Shc and that this association mediates its downregulation of insulin mitogenesis.
Because the ␤-subunit of the insulin receptor does not directly bind to CEACAM1 (38), its detection in the CEACAM1/ GST-Shc precipitate suggests that Shc binding to CEACAM1 does not interfere with its binding to the receptor. Instead, it mediates the indirect association between the receptor and CEACAM1. This is consistent with Shc binding to phosphorylated Tyr 488 of CEACAM1 at its SH2 domain and to Tyr 960 in the juxtamembrane domain of the insulin receptor at its PTB domain. Thus, Shc may function as one of the adaptor proteins that mediate the indirect association between CEACAM1 and the insulin receptor (38,39).
Like CEACAM1, cadherins (17,48) and Gab2 (66) interact with Shc at its SH2 domain. Epidermal growth factor and interleukin-3 receptors bind to both the SH2 and PTB domains of Shc (49,50). The expression level of these signaling molecules in NIH 3T3 fibroblasts is not significant. Thus, overexpressing Shc SH2 domain in these cells is not expected to alter their signaling pathways, consistent with the observation that overexpressing Shc SH2 domain in NIH 3T3 cells transfected with insulin receptors alone did not alter their growth (Fig.  5A). This is in agreement with results obtained by microinjecting Rat1 fibroblasts with this domain (51,52). Thus, the cellular changes observed upon overexpressing the Shc SH2 domain in CEACAM1-transfected NIH 3T3 cells would be chiefly attributed to the ability of this domain to specifically alter CEACAM1 binding to endogenous Shc proteins. Restoring FIG. 6. Down-regulation of the Ras/ MAP kinase mitogenesis pathway. A, NIH 3T3 cells expressing WT IR or in addition to CEACAM1-L were transfected with the SH2 domain of Shc as described in the legend to Fig. 5. Following treatment with insulin for 0 -90 min, cells were lysed in SDS buffer, and proteins were analyzed by 10% SDS-PAGE, immunoblotting with phospho-MAP kinase antibody, and detection by the LumiGlo detection system (panel i). To account for the amount of Erk1 (p44 MAP kinase) and Erk2 (p42 MAP kinase) in the immunopellet, proteins were re-immunoblotted with a p44/p42 MAP kinase antibody (panel ii). Blots were scanned, and the density of bands relative to time 0 of insulin treatment was plotted against time as measure of ERK activity (panel iii). This represents at least three experiments performed on two different clones. B, similar co-immunoprecipitation experiments as those described in the legend to Fig. 1 were performed on P IR and P IR/CC1 cells stably transfected with SH2-Shc (44 and 4, respectively). In panel i, Grb2 interaction with Shc was examined in the absence (Ϫ) or presence (ϩ) of insulin. In panels ii and iii, proteins in panel i were reprobed with ␣-Grb2 and ␣-actin antibodies to account for the amount of Grb2 and proteins in the immunopellets, respectively. In panels iv-vi, Grb2 interaction with IRS-1 was examined. This represents at least three experiments. C, cells expressing IR alone or with CEACAM1-L were insulin-treated for 0 -20 min and lysed. Proteins were then immunoprecipitated (Ip) with ␣-Shc or ␣ϪIRS-1 and immunoblotted with ␣-phospho-Tyr antibody. Bands were scanned and their density relative to time 0 of insulin treatment was calculated and plotted against duration of insulin treatment (panels iii and iv). This represents at least three experiments. growth of CEACAM1-transfected cells by overexpressing the SH2 domain of Shc (Fig. 5A) suggests that Shc binding mediates the down-regulatory effect of CEACAM1 on cell growth in response to insulin.
We herein report that CEACAM1 binding to Shc down-regulates the Ras/MAP kinase pathway by sequestering Shc and decreasing the efficiency of Grb2 coupling to the insulin receptor (Fig. 6A). Additionally, CEACAM1 binding to Shc was correlated with decreased IRS-1 phosphorylation (Fig. 6C). Because phosphorylation of Shc and IRS proteins is regulated by Tyr 960 in the juxtamembrane domain of the insulin receptor, it is conceivable that the two proteins compete for phosphorylation by the receptor, as suggested by previous observations in Rat1 fibroblasts (53). Because the cytoplasmic tail of CEACAM1 is short (71 aa), its binding to Shc may prolong the localization of this otherwise perinuclear protein in the vicinity of the plasma membrane and increase its ability to compete with IRS-1 for Tyr 960 in the insulin receptor. This reduces IRS-1 phosphorylation (Fig. 6C) and its ability to couple Grb2 to the insulin receptor and activate the Ras/MAP kinase pathway (Fig. 6B) in addition to decreasing the association between IRS-1 and the p85 subunit of PI-3 kinase and down-regulating the PI-3 kinase/Akt pathway (Fig. 7). Additionally, by recruiting Shc to the vicinity of the membrane, CEACAM1 enhances the availability of Shc to compete with the p85 subunit of PI-3 kinase for its binding to Tyr 608 in IRS-1, the main site of p85 binding to IRS-1 (54). This would down-regulate the PI-3 kinase/Akt pathway (55). Similarly, Shc competes with IRS-1 for binding at a single residue in Ret tyrosine kinase receptor to down-regulate Ret-mediated activation of the PI-3 kinase/Akt pathway (56). Even though the Ras/MAP kinase pathway constitutes the main mechanism of cell proliferation in response to insulin, the PI-3 kinase pathway also plays an important role in this event (57,58). Thus, the association between CEACAM1 and Shc appears to alter the two pathways leading to cell growth in response to insulin.
We have shown previously that co-transfecting NIH 3T3 cells with CEACAM1 decreased thymidine uptake as compared with cells expressing insulin receptors alone (33). Thus, CEACAM1 expression is correlated with decreased DNA synthesis and slower cell cycle progression. Because CEACAM1 expression down-regulates the PI-3 kinase/Akt pathway (Fig. 7) that plays a key role in mediating the anti-apoptotic effect of insulin, it is possible that CEACAM1 expression is also associated with increased cell death. This possibility warrants further investigation.
It has been suggested that CEACAM1 acts as a tumor suppressor in tissues of epithelial origin and that its intracellular domain is required for this function (40). In these studies, we observed that CEACAM1 expression decreased colonization of NIH 3T3 cells in soft agar and that overexpressing the Shc SH2 domain reversed this effect (Fig. 5C). Moreover, CEACAM1 appears to bind specifically to p46/p52 Shc that are able to transform NIH 3T3 cells in vitro (11). This suggests that Shc binding may be implicated in the molecular basis of the tumor suppression function of CEACAM1.
Like CEACAM1, other cell adhesion molecules such as cadherins bind the SH2 domain of Shc at phosphorylated tyrosines in their intracellular tail (17,48). Both cadherins and CEACAM1 mediate homophilic cell-cell adhesion even though cadherins require calcium for this function and CEACAM1 does not (16,36). Interestingly, the cytoplasmic tail of cadherins is linked to actin filaments via catenins (59). Actin binding to the cadherin-catenin complex and Shc interaction with cadherins promote cell adhesiveness (17,59). The cytoplasmic tail of CEACAM1 associates with the actin cytoskeleton through yet unidentified molecules that do not included catenins (37). Because Shc association regulates the adhesion FIG. 7. Down-regulation of the PI-3 kinase/Akt pathway by CEACAM1. A, cells expressing IR alone (left) or in addition to CEACAM1 (right) were treated with insulin for 0 -20 min prior to lysis. 300 and 450 g of homogenates derived from IR and IR/CEACAM1 (right), respectively, were immunoprecipitated (Ip) with ␣-IRS-1 antibody, analyzed by SDS-PAGE, and immunoblotted (Ib) first with ␣-p85 PI-3 kinase antibody (panel i) and then with ␣-IRS-1 antibody (panel ii) to examine IRS-1/PI-3 kinase interaction. Proteins were reprobed with ␣-actin antibody to account for the amount of proteins (panel iii). In panel iv, PI-3 kinase activity was assayed as described in the text, and the density of the spots was measured and plotted against duration of insulin treatment (panel v). This represents at least three experiments. B, Akt phosphorylation was assayed as described in the text. Briefly, cells expressing IR alone (left) or in addition to CEACAM1 (right) were treated with insulin for 0 -60 min prior to lysis in SDS buffer and analysis by 10% SDS-PAGE. Phosphorylated Akt were then detected by immunoblotting with ␣-phospho-Ser 473 Akt antibody and the LumiGlo detection system (panel i) prior to reprobing with ␣-Akt antibody (panel ii). This represents at least three experiments. function of cadherins, it is reasonable to speculate that it is involved in the complex formation between CEACAM1 and actin filaments and that this association mediates the tumor suppression function of CEACAM1.
Down-regulation of the Ras/MAP kinase pathway by other signaling molecules has been reported. These include SHIP2 in skeletal muscle (60). However, the basic mechanism of the down-regulatory effect of SHIP on the mitogenic action of insulin involves competition with Grb2 for binding phosphorylated Tyr 317 in the CH1 domain of Shc (49). Because CEACAM1 is expressed in liver and SHIP2 in skeletal muscle, the different mechanisms invoked by these molecules to reduce Shc coupling of Grb2 to the insulin receptor suggests that the effect of Shc is tightly regulated by multiple tissue-specific signaling mechanisms.
CEACAM1 also associates with SHP-2 phosphatase (61). Like Shc, SHP-2 couples Grb2 to the insulin receptor, albeit at much lower extent. Thus, it is possible that similarly to Shc, binding of CEACAM1 sequesters SHP-2 and down-regulates cell growth and proliferation in response to insulin.
In these studies, we propose that CEACAM1 provides an alternative signaling pathway that modulates the biologic action of insulin. An extension of our hypothesis is that abnormal CEACAM1 expression is associated with abnormal growth and development. Further studies are required to explore this possibility.