MUC1 Oncoprotein Functions in Activation of Fibroblast Growth Factor Receptor Signaling

The MUC1 integral membrane protein is expressed at the apical borders of normal secretory epithelial cells and aberrantly at high levels over the entire surface of carcinoma cells (1). MUC1 is translated as a single polypeptide and undergoes autocleavage into two subunits that form a heterodimeric complex (2-4). The MUC1 NH₂-terminal ectodomain (MUC1 N-ter, MUC1-N) consists of variable numbers of 20 amino acid tandem repeats, which are modified by O-linked glycans (5, 6). The MUC1 COOH-terminal subunit (MUC1 C-ter, MUC1-C) includes a 58-amino acid extracellular domain, a transmembrane domain, and a 72-amino acid cytoplasmic tail (7). The >250-kDa MUC1-N ectodomain associates with the ∼25-kDa MUC1-C as a heterodimer or, in response to certain stimuli, is released from the cell surface (2). Importantly, overexpression of MUC1 confers transformation and blocks the apoptotic response to anticancer agents (8-12). MUC1 interacts with members of the ErbB family of receptor tyrosine kinases (13-15). Stimulation of cells with the epidermal growth factor (EGF) induces tyrosine phosphorylation of the MUC1 cytoplasmic domain (MUC1-CD) and binding of MUC1 to the Wnt effector β-catenin (13, 16-18). Moreover, heregulin stimulates interactions between MUC1 and ErbB2, ErbB3, and ErbB4 and binding of MUC1-C to β-catenin and the related γ-catenin (15). Heregulin also targets MUC1-C, but not MUC1-N, to mitochondria (10). These findings have indicated that MUC1-C functions in transducing signals from activated ErbB receptors to the interior of the cell. MUC1 is not known to interact with other cell surface receptors.

The fibroblast growth factor (FGF) receptor (FGFR) family of receptor tyrosine kinases consists of four related members that have been designated FGFR1, FGFR2, FGFR3, and FGFR4. The FGFRs contain three extracellular immunoglobulin (Ig) domains, a transmembrane region, and an intracellular cytoplasmic tail with a divided tyrosine kinase domain (19). Activation of FGFRs by the FGF family of ligands results in receptor dimerization and transphosphorylation of the intracellular domains on tyrosine (19). Mutations in the genes encoding the FGFRs confer several different disorders that affect skeletal development (20). Mutations and translocations of the FGFRs have also been associated with human cancers (21-25). In particular, mutations that result in aberrant activation of FGFR3 have been found in bladder carcinomas, cervical carcinomas, and multiple myeloma (26-28). The t(4;14) translocation that results in FGFR3 activation in multiple myeloma cells is associated with increased proliferation (27-31). Moreover, silencing FGFR3 induces apoptosis of multiple myeloma cells (32). Overexpression of FGFR3 has also been linked to the development of bladder cancer (33). FGFR3 has been shown to signal activation of signal transducers and activators of transcription 1, 3, and 5 (34-37). Otherwise, little is known about how FGFR3 may contribute to transformation (38).

The present results show that MUC1 associates with FGFR3 and that this interaction is increased by FGF1. We show that FGF1 induces phosphorylation of MUC1-C on a YEKV motif in the cytoplasmic tail and thereby binding of MUC1 to β-catenin or the heat shock protein (HSP) 90 chaperone. The results also show that FGF1 induces targeting of MUC1-C to the nucleus with β-catenin and to mitochondria by a HSP90-dependent mechanism.

To determine if FGFR3 forms a complex with MUC1, anti-FGFR3 immunoprecipitates from human ZR-75-1 breast cancer cell lysates were analyzed by immunoblotting with anti-MUC1. The results show that MUC1 coprecipitates with FGFR3 (Fig. 1A, left). Analysis of the immunoprecipitates with anti-FGFR3 confirmed precipitation of FGFR3 (Fig. 1A, left). As a control, there was no detectable MUC1 or FGFR3 in precipitates prepared with nonimmune rabbit IgG (Fig. 1A, left). In the reciprocal experiment, analysis of anti-MUC1 immunoprecipitates with anti-FGFR3 confirmed that MUC1 associates with FGFR3 (Fig. 1A, right). To assess whether activation of FGFR3 affects the interaction with MUC1, ZR-75-1 cells were stimulated with FGF1. Immunoblot analysis of anti-MUC1 immunoprecipitates with anti-FGFR3 showed that FGF1 increases the association between FGFR3 and MUC1 (Fig. 1B). To confirm this interaction, HCT116 colon cancer cells, which are null for endogenous MUC1 expression, were stably transfected to express an empty vector, MUC1, or MUC1(Y46F) mutant. Tyr⁴⁶ is the residue within the YEKV motif in the MUC1-CD that is phosphorylated by c-Src (18). FGF1 stimulation was associated with binding of FGFR3 and MUC1 in HCT116/MUC1, but not in HCT116/vector or HCT116/MUC1(Y46F), cells (Fig. 1C), suggesting that the MUC1-CD is functional in this interaction. Consequently, we overexpressed green fluorescent protein (GFP)-tagged MUC1-CD or MUC1-CD fragments in 293 cells. Immunoblot analysis of anti-GFP immunoprecipitates with anti-FGFR3 showed that MUC1-CD associates with FGFR3 (Fig. 1D). MUC1-CD(20-72) and MUC1-CD(46-72), but not MUC1-CD(1-45), also formed complexes with FGFR3 (Fig. 1D), indicating that MUC1-CD amino acids 46 to 72 are sufficient for the interaction. These findings indicate that (a) MUC1 associates with FGFR3, (b) activation of FGFR3 increases the association with MUC1, and (c) MUC1-CD is of importance in forming complexes with FGFR3.

EGF stimulation is associated with phosphorylation of MUC1 on Tyr⁴⁶ (13). To determine if FGF1 induces a similar response, lysates from ZR-75-1 breast cancer cells stably transfected to express a MUC1 small interfering RNA (siRNA) or, as a control, the empty vector (10) were analyzed by immunoblotting with an antibody against phosphorylated MUC1-Tyr⁴⁶ (Fig. 2A). A low level of MUC1-Tyr⁴⁶ phosphorylation was detectable in control ZR-75-1/vector cell lysates (Fig. 2A). By contrast, there was no detectable anti-phosphorylated MUC1-Tyr⁴⁶ reactivity in cells silenced for MUC1 (Fig. 2A). Stimulation with FGF1 for 10 minutes was associated with a substantial increase in anti-phosphorylated MUC1-Tyr⁴⁶ reactivity (Fig. 2B, left). Quantitation of the intensity of the signals by densitometric scanning showed over a 2-fold increase at 10 to 30 minutes of FGF1 stimulation in three separate experiments (Fig. 2B, right). Moreover, stimulation with FGF1 for 1 hour was associated with a return toward baseline levels of MUC1-Tyr⁴⁶ phosphorylation (Fig. 2B, right). Previous work showed that MUC1-Tyr⁴⁶ is phosphorylated by c-Src in heregulin-stimulated cells (39). In the present studies, treatment of the ZR-75-1 cells with 4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrazolo(3,4-d)pyrimidine (PP2), an inhibitor of c-Src, blocked FGF1-induced phosphorylation of MUC1 on Tyr⁴⁶, indicating that this response is mediated by a c-Src-dependent mechanism (Fig. 2B). As a control, there was no detectable effect of FGF1 stimulation on MUC1-C levels (Fig. 2B, left). We previously reported that EGF stimulation of HCT116/MUC1 cells induces phosphorylation of MUC1 on Tyr⁴⁶ (18). Similar results were obtained when lysates from EGF-stimulated ZR-75-1 cells were immunoblotted with anti-phosphorylated MUC1-Tyr⁴⁶ (Fig. 2C, left). Quantitation of the intensity of signals obtained in three experiments showed that EGF induces similar levels of MUC1-Tyr⁴⁶ phosphorylation as found in the FGF1-stimulated cells (Fig. 2C, right). Moreover, like FGF1, the response to EGF was sensitive to PP2 treatment (Fig. 2C). Phosphorylation of MUC1 on Tyr⁴⁶ increases the interaction between MUC1 and β-catenin (13, 18). Consistent with those findings and FGF1-induced phosphorylation of Tyr⁴⁶, FGF1 stimulation of ZR-75-1 cells was associated with increases in MUC1-β-catenin complexes (Fig. 2D). In HCT116 cells, FGF1 also induced binding of MUC1 and β-catenin (Fig. 2E). Moreover, this response was abrogated in HCT116 cells expressing MUC1 with a Y46F mutation (Fig. 2E). These findings indicate that FGF1 stimulation is associated with increased (a) phosphorylation of MUC1 on Tyr⁴⁶ and (b) binding of MUC1 and β-catenin.

MUC1 accumulates in the cytosol of transformed cells (1, 39-42). In concert with these findings, MUC1-C was detectable in the cytosol of ZR-75-1 cells (Fig. 3A). Moreover, stimulation of ZR-75-1 cells with FGF1 was associated with an increase in cytosolic MUC1-C levels (Fig. 3A, left). Quantitation of the signals from three experiments showed an ∼2-fold increase in MUC1-C in the cytosol (Fig. 3A, right). Purity of the cytosolic preparations was confirmed by immunoblotting with antibodies against the cell membrane–associated platelet-derived growth factor receptor, endoplasmic reticulum–associated BAP31 (43), and the mitochondria-associated Tom20 proteins (Fig. 3A). A similar increase in cytosolic MUC1-C levels was observed when HCT116/MUC1 cells were stimulated with FGF1 (Fig. 3B), indicating that this response is not cell type specific. Other studies have shown that MUC1-C, but not MUC1-N, is trafficked to the nucleus (8, 9, 11, 44, 45). Analysis of purified nuclei from ZR-75-1 cells showed that FGF1 increases MUC1-C levels (Fig. 3C). Consistent with recent findings that β-catenin is stabilized by direct binding to MUC1 and that MUC1 targets β-catenin to the nucleus (46), nuclear levels of β-catenin were also increased over 2-fold in the response to FGF1 stimulation (Fig. 3C). Targeting of MUC1-C to the nucleus by FGF1 was confirmed in HCT116/MUC1 cells (Fig. 3D). However, in concert with previous results (46), the FGF1-induced interaction between MUC1 and β-catenin had no effect on nuclear β-catenin levels in HCT116 cells (Fig. 3D), which express a mutant β-catenin with deletion of Ser⁴⁵. This mutation abrogates the glycogen synthase kinase 3β phosphorylation-dependent binding of β-catenin with β-TrCP and thereby confers deregulated stability of β-catenin (47-50). These findings indicate that FGF1 regulates targeting of MUC1 to the cytosol and nucleus.

Recent studies showed that heregulin targets MUC1-C to mitochondria (10) and that this response is mediated by the HSP70 and HSP90 chaperones (39). As found with heregulin (39), FGF1 stimulation had little if any effect on binding of MUC1-C to HSP70 in HCT116/MUC1 (Fig. 4A) and ZR-75-1 (data not shown) cells. Stimulation of HCT116/MUC1 cells with FGF1 was, however, associated with increased binding of MUC1-C to HSP90, and this response was similar to that observed with heregulin (Fig. 4B). FGF1 also induced binding of MUC1-C to HSP90 in ZR-75-1 cells (Fig. 4C). In heregulin-stimulated cells, binding of MUC1-C to HSP90 is mediated by phosphorylation of MUC1 on Tyr⁴⁶ (39). FGF1-induced interaction between MUC1-C and HSP90 was attenuated by expression of MUC1 with the Y46F mutation (Fig. 4D). These findings indicate that binding of MUC1-C to HSP90 is induced in the response of cells to FGF1 stimulation.

The results thus indicate that FGF1-induced phosphorylation of MUC1 on Tyr⁴⁶ increases binding of β-catenin and HSP90. We therefore asked whether binding of MUC1 to β-catenin and HSP90 is a mutually exclusive event. Previous work has shown that (a) β-catenin binds to a serine-rich motif (SRM) in the COOH-terminal region of the MUC1-CD, (b) this binding is increased by c-Src phosphorylation of MUC1-CD, and (c) mutation of the SRM abrogates the interaction (13, 16-18, 46). Other studies have shown that HSP70 binds to the MUC1-CD COOH-terminal region (39). The present results show that, in contrast to β-catenin, direct binding of MUC1-CD to HSP70 is not abrogated by mutation of the SRM (Fig. 5A). As shown previously (39), direct binding to HSP90 was dependent on c-Src phosphorylation of MUC1-CD (Fig. 5B). Moreover, there was no detectable HSP90 binding to c-Src-phosphorylated MUC1-CD with a mutated SRM (Fig. 5B), indicating that β-catenin and HSP90 may both interact with this region. In competition experiments, incubation of MUC1-CD with HSP70 and then increasing amounts of β-catenin was associated with a progressive decrease in the interaction between MUC1-CD and HSP70 (Fig. 5C). In addition, β-catenin was more effective in competing with HSP90 for binding to MUC1-CD that had been phosphorylated by c-Src (Fig. 5D). These findings indicate that MUC1 forms mutually exclusive complexes with β-catenin and HSP70/HSP90.

To further assess FGF1-induced binding of MUC1-C to HSP90, we treated cells with geldanamycin, an inhibitor of HSP90 function (51). Geldanamycin had no effect on the interaction between MUC1 and HSP70 (Fig. 6A) but blocked the FGF1-induced binding of MUC1 and HSP90 (Fig. 6B). The HSP70/HSP90 complex functions in the delivery of client proteins to the mitochondrial outer membrane (52). In this regard, FGF1 stimulation was associated with increased targeting of MUC1 to mitochondria (Fig. 6C). Immunoblot analysis with antibodies against the nuclear proliferating cell nuclear antigen and cytosolic IκBα proteins further indicated that the mitochondria were not significantly contaminated with these fractions (Fig. 6C). In addition and consistent with involvement of HSP90, treatment with geldanamycin blocked FGF1-induced targeting of MUC1 to mitochondria (Fig. 6C). To confirm these results, we treated cells with 17-(allylamino)-17-demethoxygeldanamycin (17-AAG), another inhibitor of HSP90 (53). As found with geldanamycin, FGF1-induced increases in mitochondrial MUC1-C were blocked by 17-AAG (Fig. 6D). These findings indicate that FGF1 targets MUC1 to mitochondria by a HSP90-dependent mechanism.

MUC1 is expressed at the cell membrane as a heterodimer of the MUC1-N and MUC1-C subunits. Under physiologic conditions, MUC1 localizes to the apical borders of normal secretory epithelia (1). However, with transformation and loss of polarity, MUC1 is overexpressed throughout the entire cell membrane of carcinoma cells (1), allowing MUC1 to interact with receptors that are normally restricted to the lateral and basal borders. In this regard, previous work has shown that MUC1 associates with members of the ErbB family of receptor tyrosine kinases in human breast cancer cells (13-15). Moreover, stimulation of cells with EGF or heregulin increased ErbB receptor-MUC1 complexes and tyrosine phosphorylation of the MUC1-CD (13-15). The present studies show that MUC1 also associates with FGFR3, another member of the receptor tyrosine kinase family. Importantly, activation of FGFR3 has been linked to the development of human malignancies; however, little is known about how FGFR3 contributes to transformation (38). FGFR3 is activated by its ligand FGF1 (19). As found with activation of ErbB receptors, stimulation of FGFR3 with FGF1 was associated with increased binding of FGFR3 and MUC1. In addition, stimulation of cells with FGF1 was associated with phosphorylation of MUC1 on Tyr⁴⁶ in the cytoplasmic domain. As found with EGF, pretreatment of cells with PP2 blocked FGF1-induced phosphorylation of MUC1 on Tyr⁴⁶, indicating that this response to both growth factors is mediated by a c-Src-dependent mechanism. These findings indicate that, in addition to the ErbB receptors, MUC1 interacts with FGFR3 and that MUC1 may be of importance to intracellular signaling in cancers with FGFR3 mutations or overexpression.

The MUC1-Tyr⁴⁶ site is located just upstream to a SRM that binds to β-catenin (16). Previous work on cells stimulated with EGF showed that phosphorylation of MUC1 on Tyr⁴⁶ induces binding of MUC1 and β-catenin (13). In concert with these results, the present studies show that FGF1-induced phosphorylation of MUC1 on Tyr⁴⁶ is also associated with increased binding of MUC1 and β-catenin. Moreover, expression of MUC1 with a Y46F mutation attenuated FGF1-induced formation of MUC1-β-catenin complexes. The MUC1 SRM binds directly to the β-catenin armadillo repeats and blocks glycogen synthase kinase 3β-mediated phosphorylation and degradation of β-catenin (46). MUC1-induced stabilization of β-catenin is also associated with increases in nuclear β-catenin and activation of Wnt target genes (46). We previously showed that MUC1 increases β-catenin levels in ZR-75-1 cells, but not in HCT116 cells, which express a mutant and thereby stable form of β-catenin (46). In concert with these results, FGF1-induced binding of MUC1 and β-catenin was associated with increased targeting of β-catenin to the nucleus of ZR-75-1 cells. By contrast, FGF1-induced binding of MUC1 and β-catenin had no effect on nuclear β-catenin levels in HCT116 cells. These findings indicate that MUC1 functions in integrating FGF1-induced signaling to β-catenin and thereby the Wnt pathway.

Recent work has shown that MUC1 also forms intracellular complexes with HSP70 and HSP90 (39). The MUC1-CD binds directly to HSP70 (39). In addition, binding of MUC1 to HSP90 was found to be dependent on phosphorylation of MUC1 on Tyr⁴⁶ (39). The present studies confirm the constitutive interaction between MUC1 and HSP70 and show that FGF1 stimulation has little if any effect on this interaction. We also found that FGF1-induced phosphorylation of MUC1 on Tyr⁴⁶ stimulates binding of MUC1 to HSP90. Moreover, FGF1 increased the interaction between MUC1 and HSP90 by a mechanism that was attenuated by expressing MUC1 with the Y46F mutation. Geldanamycin, an inhibitor of the ATP-driven HSP90 chaperone cycle (51), also attenuated FGF1-induced binding of MUC1 to HSP90. Binding of HSP90 is associated with stabilization of certain client proteins (53). However, the finding that geldanamycin has no effect on MUC1 levels indicates that binding of this chaperone to MUC1 is not involved in the regulation of MUC1 stability. In this regard, the available evidence indicates that binding of MUC1 to complexes of HSP70 and HSP90 is responsible for delivery of MUC1 to the mitochondrial outer membrane (39).

Our results indicate that FGF1-induced phosphorylation of MUC1 on Tyr⁴⁶ increases binding of MUC1 to both β-catenin and HSP90. Whereas β-catenin binds to the SRM in the MUC1-CD (16, 46), the region responsible for the interaction with HSP90 has not been clear (39). Direct binding studies in the present work show that mutation of the SRM has no effect on binding of MUC1-CD to HSP70 but abrogates the interaction with HSP90. How HSP90 interacts with different client proteins is not precisely known; however, the available evidence indicates that HSP90 recognizes a surface structure rather than an amino acid sequence (54). Thus, our results suggest that the SRM may contribute to a surface structure on MUC1-CD that confers binding to HSP90. Consistent with involvement of the SRM, our studies further show that β-catenin competes with HSP90 for binding to MUC1-CD. In addition, β-catenin was effective, but to a lesser extent, in displacing HSP70 from MUC1-CD. These findings indicate that, following FGF1 stimulation and phosphorylation of MUC1 on Tyr⁴⁶, MUC1 can interact with either β-catenin or the HSP70/HSP90 complex (Fig. 7). The extent of MUC1 binding to β-catenin or HSP70/HSP90 complexes may be dictated by the relative intracellular levels of these proteins and/or other signals that are dependent on cell context. The findings also support a model in which intracellular trafficking of MUC1-C is in turn dictated by forming these mutually exclusive complexes (Fig. 7).

The MUC1-C, but not the MUC1-N, subunit trafficks to the nucleus (8-10, 15, 44-46, 55) and mitochondria (10, 39). The 20- to 25-kDa MUC1-C subunit includes the extracellular, transmembrane, and cytoplasmic domains (10). MUC1-C is also detectable as 17- and 15-kDa species (10); however, the precise structures of these forms are presently not known. The present studies indicate that FGF1 stimulation is associated with targeting of MUC1-C and β-catenin to the nucleus. The mechanism responsible for nuclear import of MUC1-C is not known, although MUC1-C associates with importin α and β, which function in transporting proteins to the nuclear pore.¹

FGF1-induced targeting of MUC1-C to mitochondria was attenuated by geldanamycin, consistent with a mechanism in which HSP70/HSP90 complexes deliver client proteins to the mitochondrial outer membrane (52). By contrast, geldanamycin had no detectable effect on FGF1-induced targeting of MUC1-C to the nucleus (data not shown). These results collectively indicate that FGF1 activates two pathways in which MUC1-C is targeted to (a) the nucleus by a HSP90-independent mechanism and (b) mitochondria by a HSP90-dependent mechanism (Fig. 7). Delivery of MUC1-C to mitochondria is associated with integration of MUC1-C into the mitochondrial outer membrane (10, 39). The precise function of mitochondrial MUC1-C is not clear; however, localization of MUC1-C to the mitochondrial outer membrane is associated with attenuation of stress-induced (a) loss of mitochondrial transmembrane potential, (b) release of mitochondrial apoptogenic factors, (c) activation of caspase-3, and (d) cell death (10, 58, 59). These findings thus support a model in which the FGF1-induced interaction between MUC1-C and HSP70/HSP90 complexes transduces signals from the cell membrane to mitochondria that attenuate activation of the intrinsic apoptotic pathway in the response to stress (Fig. 7).

Human ZR-75-1, ZR-75-1/vector, and ZR-75-1/MUC1siRNA breast cancer cells (10) were grown in RPMI 1640 containing 10% heat-inactivated fetal bovine serum, 100 units/mL penicillin, 100 μg/mL streptomycin, and 2 mmol/L l-glutamine. The ZR-75-1/MUC1siRNA cells were stably infected with a retrovirus expressing a siRNA that targets the 5′-GGTACCATCAATGTCCACG-3′ sequence encoding amino acids in the MUC1-C extracellular domain (10). The ZR-75-1/vector cells were infected with the empty retrovirus (10). HCT116/vector, HCT116/MUC1, and HCT116/MUC1(Y46F) cells (10) were cultured in DMEM/F12 with 10% heat-inactivated fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin. Human 293 embryonic kidney cells were grown in DMEM containing 10% heat-inactivated fetal bovine serum, antibiotics, and l-glutamine. Cells were cultured overnight in medium with 0.1% heat-inactivated fetal bovine serum and then stimulated with 30 ng/mL FGF1 (R&D Systems, Minneapolis, MN) in the presence of 7.5 μg/mL heparin (Sigma-Aldrich, St. Louis, MO), 20 ng/mL EGF (EMD Biosciences, La Jolla, CA), or 20 ng/mL heregulin (EMD Biosciences) at 37°C. Cells were also treated with 10 μmol/L PP2 (EMD Biosciences), 1 μmol/L geldanamycin (EMD Biosciences), or 17-AAG (EMD Biosciences).

Cell lysates prepared as described (10) were incubated with anti-FGFR3 (Sigma-Aldrich), anti-MUC1-N (monoclonal antibody DF3; ref. 1), anti-GFP (Clontech, Palo Alto, CA), anti-HSP70 (BD Biosciences, Palo Alto, CA), anti-HSP90 (BD Biosciences), or nonimmune IgG (Santa Cruz Biotechnology, Santa Cruz, CA) for 2 hours at 4°C. Immune complexes were subjected to immunoblotting with anti-MUC1-N, anti-FGFR3, anti-GFP, anti-MUC1-C (Ab5, NeoMarkers, Fremont, CA), anti-β-catenin (Zymed Laboratories, Carlsbad, CA), anti-HSP70, and anti-HSP90. Lysates not subjected to precipitation were immunoblotted with anti-phosphorylated MUC1-Tyr⁴⁶ (Cell Signaling Technology, Danvers, MA), anti-MUC1-C, anti-β-actin (Sigma-Aldrich), anti-platelet-derived growth factor receptor (EMD Biosciences), anti-BAP31 (Abcam, Cambridge, MA), anti-Tom20 (BD Biosciences), anti-IκBα (Santa Cruz Biotechnology), anti-lamin B (EMD Biosciences), anti-HSP60 (Stressgen Bioreagents, Victoria, B.C., Canada), and anti-proliferating cell nuclear antigen (EMD Biosciences). Reactivity was detected with horseradish peroxidase–conjugated secondary antibodies and chemiluminescence (PerkinElmer Life Sciences, Wellesley, MA). Intensity of the signals was determined by densitometric scanning. Statistical significance was determined by Student's t test.

Sequences encoding MUC1-CD, MUC1-CD(1-45), or MUC1-CD(46-72) were amplified by PCR and cloned into the pEGFP-C1 vector (Clontech). Transient transfections were done in the presence of LipofectAMINE (Invitrogen Life Technologies, Carlsbad, CA).

Cytosolic and nuclear fractions were purified as described (60, 61). Mitochondria were prepared as described (10).

Purified glutathione S-transferase (GST), GST-MUC1-CD, and GST-MUC1-CD(mSRM) in which the SRM SAGNGGSSLS was mutated to AAGNGGAAAA (39, 46) were bound to glutathione beads and incubated with recombinant human HSP70 or HSP90 (Stressgen Bioreagents) for 1 hour at 4°C. In certain experiments, the purified GST fusion proteins bound to glutathione beads were incubated with or without 50 ng c-Src (Upstate Biotechnology, Lake Placid, NY) in the presence of 200 μmol/L ATP for 30 minutes at 30°C. HSP90 was then added and the reaction was incubated for 1 hour at 4°C. For competition studies, HSP70 bound to GST-MUC1-CD or HSP90 bound to c-Src-phosphorylated GST-MUC1-CD was incubated with increasing amounts of β-catenin for 2 hours at 4°C. Precipitated proteins were subjected to immunoblot analysis.

We thank Dr. Q. Ge (Cell Signaling Technology) for the anti-phosphorylated MUC1-Tyr⁴⁶ antibody and Kamal Chauhan for technical support.