The Phosphoprotein Stard10 Is Overexpressed in Breast Cancer and Cooperates with Erbb Receptors in Cellular Transformation

[CANCER RESEARCH 64, 3538–3544, May 15, 2004] The Phosphoprotein StarD10 Is Overexpressed in Breast Cancer and Cooperates with ErbB Receptors in Cellular Transformation

Monilola A. Olayioye,1 Peter Hoffmann,2 Thomas Pomorski,3 Jane Armes,4 Richard J. Simpson,5 Bruce E. Kemp,2 Geoffrey J. Lindeman,1 and Jane E. Visvader1 1The Walter and Eliza Hall Institute of Medical Research and Bone Marrow Research Laboratories, Royal Melbourne Hospital, Parkville, Victoria, Australia; 2St. Vincent’s Institute of Medical Research, St. Vincent’s Hospital, Victoria, Australia; 3Humboldt-University Berlin, Institute of Biology and Biophysics, Berlin, Germany; 4Molecular Pathology Laboratory, Victorian Breast Cancer Research Consortium, Department of Pathology, University of Melbourne, Victoria, Australia; and 5Joint Proteomics Laboratory, Ludwig Institute for Cancer Research and The Walter and Eliza Hall Institute of Medical Research, Melbourne, Victoria, Australia

ABSTRACT tion of the phosphatidylinositol 3-kinase (PI3K) pathway (15, 16). Signaling through PI3K plays an important role in cellular survival by We have identified that StarD10, a member of the START protein phosphorylating and inactivating growth-inhibitory and proapoptotic family, is overexpressed in both mouse and human breast tumors. proteins, including the FOXO transcription factors (17). In addition, StarD10 was initially discovered on the basis of its cross-reactivity with a phosphoserine-specific antibody in mammary tumors from Neu/ErbB2 the ErbB2 and ErbB3 receptors recruit the adaptor proteins Shc and transgenic mice and subsequently isolated from SKBR3 human breast Grb2 (16, 18, 19), resulting in stimulation of the Ras-Raf-mitogen- carcinoma cells using a multistep biochemical purification strategy. We activated protein kinase pathway (20, 21), which has been implicated have shown that StarD10 is capable of binding lipids. StarD10 was found in mammary tumor progression (22, 23). Furthermore, Src kinase to be overexpressed in 35% of primary breast carcinomas and 64% of activity is enhanced in both mammary tumors from Neu transgenic human breast cancer cell lines, correlating with their ErbB2/Her2 status. mice (24) and human breast tumors (25, 26), and Src cooperates with Coexpression of StarD10 with ErbB1/epidermal growth factor receptor in ErbB1 in the transformation of mouse fibroblasts (27). Thus, aberrant murine fibroblasts enhanced anchorage-independent growth in soft agar, expression of ErbB receptors triggers the activation of multiple down- providing evidence for functional cooperation between StarD10 and ErbB stream effectors, in addition to synergizing with other proto-onco- receptor signaling. Taken together, these data suggest that overexpression genes to transform mammary epithelium. of this lipid-binding protein contributes to breast oncogenesis. The Neu mouse model has served as a valuable model for the identification of intracellular mediators of ErbB2-induced tumor de- INTRODUCTION velopment. Although ErbB receptor signaling has been studied exten- The ErbB family of receptor tyrosine kinases plays a critical role in sively, little is known about the deregulation of transcription factors in the pathogenesis of breast cancer. Amplification and/or overexpres- ErbB2-induced tumorigenesis. In the course of analyzing the expression of HER2/ErbB2 occurs in 20–30% of human breast cancers and sion and activation of the forkhead transcription factor FKHR/FOXO1 correlates with poor prognosis (1). Deregulated expression of HER1/ in Neu transgenic mice, we observed a novel protein band that was ErbB1/epidermal growth factor (EGF) receptor has also been ob- specifically recognized by a phospho-FKHR antibody. Interestingly, served in breast cancer, often in association with aberrant expression this Mr 35,000 protein was abundant in tumors derived from Neu of ErbB2 (2, 3). ErbB1 and ErbB2, together with HER3/ErbB3 and transgenic mice but not adjacent normal tissue. Here, we describe the HER4/ErbB4, constitute the ErbB family of receptor tyrosine kinases. isolation and characterization of this phosphoprotein, StarD10, a These receptors signal cooperatively by forming ligand-induced com- member of the START domain family of lipid-binding proteins. binations of homo- and hetero-dimers (4, 5). The complex network of StarD10 was found to be coexpressed with ErbB2 in many breast ErbB receptor-ligand combinations provides precise signaling along carcinoma lines and cooperated with the ErbB pathway in cellular pathways underlying diverse developmental processes (6–8). transformation. Finally, StarD10 overexpression was observed in a The importance of ErbB2 signaling in mammary tumorigenesis has high proportion of primary breast cancers, supporting a role for this been established using transgenic mouse models. Overexpression of lipid-binding protein in deregulated cell growth and tumorigenesis. neu, the rat homologue of ErbB2, from the mouse mammary tumor virus promoter induces focal mammary tumors which frequently MATERIALS AND METHODS metastasize to the lung (9). The long tumor latency period in these mice suggested that additional molecular events were required for Reagents. Antibodies used were polyclonal antiphospho-FKHR (S256) antibody (New England Biolabs, Beverly, MA), mouse anti-␣-tubulin mono- mammary tumors to develop. Further analysis of these tumors iden- clonal antibody (clone B-5-1-2; Sigma, St. Louis, MO), mouse anti-ErbB2 tified somatic mutations in neu, resulting in constitutive activation of monoclonal antibody (Ab-10; Neomarkers, Fremont, CA), rabbit anti-ErbB2 the receptor (10). Up-regulation of ErbB3 protein in these tumors polyclonal antibody (A0485; DAKO, Carpinteria, CA), mouse anti-Neu mono- suggested that ErbB2 synergizes with ErbB3 in oncogenesis (11). In clonal antibody (Ab-3; Calbiochem, Darmstadt, Germany), mouse anti-Flag human breast carcinoma cells that overexpress ErbB2, down-regula- monoclonal antibody (Sigma), and rabbit anti-Pctp-l polyclonal antiserum tion of ErbB3 using an artificial erbB3-specific transcriptional repres- (kindly provided by Y. Nishimune, Osaka University, Osaka, Japan). sor revealed that ErbB3 is essential for ErbB2-mediated proliferation Cell Culture and Mouse Strains. The majority of breast epithelial cell (12). The ErbB2/3 heterodimer appears to have potent mitogenic and lines was maintained in RPMI containing 10% fetal bovine serum (Common- transforming properties in vitro (13, 14) and leads to efficient activa- wealth Serum Laboratory). MCF10A cells were grown in DMEM:F12, supplemented with 10% fetal bovine serum, 10 ng/ml EGF (Sigma), 5 ␮g/ml insulin (Sigma), and 1 ␮M dexamethasone (Sigma). HER14 (NIH3T3 cells Received 11/30/03; revised 2/20/04; accepted 3/3/04. transfected with human ErbB1) was a gift from R. Daly (Garvan Institute, Grant support: The Victorian Breast Cancer Research Consortium, Inc. M. Olayioye was supported by EMBO and HFSPO fellowships. Sydney, Australia). HER14, 293T, and Bosc cell lines were grown in DMEM The costs of publication of this article were defrayed in part by the payment of page containing 10% fetal bovine serum. HER14 StarD10 transductants were gen- charges. This article must therefore be hereby marked advertisement in accordance with erated by infection with a pBabe retrovirus encoding hemagglutinin (HA)- 18 U.S.C. Section 1734 solely to indicate this fact. tagged StarD10 into Bosc packaging cells. Virus was collected, then used to Requests for reprints: Jane E. Visvader, The Walter and Eliza Hall Institute of ␮ Medical Research, 1G Royal Parade, Victoria 3050, Australia. Phone: 61-3-9342-8633; infect HER14 NIH3T3 cells, and selected with 1 g/ml puromycin (Sigma), Fax: 61-3-9347-0852; E-mail: [email protected]. thus generating HER14-StarD10 transductants. For transient transfections, 3538 STARD10 IN BREAST CANCER

293T cells were transfected with Fugene reagent (Roche, Penzberg, Germany) washed with acetone, dissolved in sample buffer, and loaded onto a large 10% according to the manufacturer’s instructions. Mouse mammary tumor virus- SDS-PAGE gel. The gel was fixed and stained with Coomassie Phast-gel Blue neu mice have been described previously (9). Nulliparous female mice were R (Amersham Pharmacia). aged for 8–12 months in standard animal facilities at The Walter and Eliza In-Gel Tryptic Digest and Mass Spectrometry. Coomassie-stained bands Hall Institute and sacrificed when tumors developed. were destained with 50 mM NH4HCO3/50% acetonitrile (1:1) and digested Cloning of StarD10. Total RNA was extracted from SKBR3 cells using with modified porcine trypsin (Promega, Madison, WI). The sample was then RNAzol (Tel-Test, Friendswood, TX). First-strand cDNA synthesis was per- concentrated on a Zip Tip ␮-C18 (Millipore) and eluted into nanospray formed, and StarD10 cDNA was amplified by PCR using primers based on the capillaries (MDS Proteomics, Charlottesville, VA) with 20–60% methanol CGI-52 sequence (accession no. AF151810; Ref. 28). StarD10 cDNA was containing 5% formic acid. Mass spectrometry was performed on a Pulsar ion cloned into Flag-pEFrPGKpuro and pBabePuro vectors for expression in quadruple Time of Flight (TOF) mass spectrometer (Applied Biosystems, mammalian cells. Foster City, CA) with a nanoelectrospray ion source (MDS Proteomics). Protein Extraction of Cells and Tissues. For cytosolic protein extraction, Product ion scans were acquired using collision energies that retained 10% of cells were lysed in hypotonic buffer [10 mM HEPES (pH 7.9), 133 mM sorbitol, the intensity of the precursor ion. Argon was used as collision gas at a recorded 0.5 mM sodium fluoride, and 0.5 mM ␤-glycerophosphate plus Complete ϫ Ϫ5 protease inhibitors; Roche], left to swell on ice for 10 min, homogenized by pressure of 4.3 10 Torr. Tandem mass spectroscopy data were searched douncing, and then centrifuged at 800 ϫ g for 10 min. The pellet was washed via the Mascot search engine (30) or sequenced de novo. with hypotonic buffer, and the supernatants were combined to yield the StarD10-Lipid Interaction by Intrinsic Fluorescence Measurement. cytosolic fraction. Whole-cell extracts were obtained by solubilizing cells in StarD10 cDNA was cloned into pGEX-6P (Amersham Pharmacia) and trans- NP40 extraction buffer (NEB) [50 mM Tris (pH 7.5), 150 mM NaCl, 1% NP40, formed into BL21 bacteria to produce a glutathione S-transferase StarD10 1mM sodium orthovanadate, 10 mM sodium fluoride, and 20 mM ␤-glycero- fusion protein. Glutathione S-transferase-StarD10 expression was induced with phosphate plus Complete protease inhibitors]. Lysates were clarified by cen- 0.5 M isopropyl-␤-D-1-thiogalactopyranoside for4hat30°C. Bacteria were trifugation at 16,000 ϫ g for 10 min. Protein lysates of mammary tissue were harvested and resuspended in PBS containing 50 ␮g/ml lysozyme, Complete obtained by grinding the tissue to a fine powder under liquid nitrogen and protease inhibitors, 10 mM sodium fluoride, and 20 mM ␤-glycerophosphate. subsequent solubilization in Triton X-100 extraction buffer (TEB) (see above, Triton was added to a final concentration of 1% before two freeze-thaw cycles substitution of NP40 with Triton X-100). Lysates were homogenized by and sonication. Glutathione S-transferase-StarD10 was purified from the clar- passing through a 21-gauge needle before clarification by centrifugation. ified lysate with glutathione resin and cleaved with Prescission protease Lysates from organs were prepared by resuspending cells in 20 mM Tris (pH according to the manufacturer’s instructions (Amersham Pharmacia). The 7.4), 135 mM NaCl, 1.5 mM MgCl2,1mM EGTA, 1% Triton X-100, and 10% purity of cleaved StarD10 protein was verified by SDS-PAGE and Coomassie glycerol with Complete protease inhibitors and processed as described for staining. Fluorescence measurements were on a Aminco Bowman spectrom- whole-cell extracts. eter series 2 (Rochester, NY). Tryptophan and tyrosine were excited at 280 nm, Immunoprecipitation and Western Blotting. For immunoprecipitations, and emission spectra were recorded from 290 to 400 nm (band widths: 4 nm). equal amounts of protein were incubated with specific antibodies for2honice. Porcine brain lipids (Avanti Polar Lipids, Alabaster) were dried under nitro- Immune complexes were collected with protein G-Sepharose (Amersham gen, and small unilamellar vesicles were prepared by sonication to yield a final Pharmacia, Buckinghamshire, United Kingdom) and washed three times with lipid concentration of 1 mM in 20 mM HEPES (pH 7.2), 100 mM NaCl, and 5 NEB (see above). For phosphorylation analysis, immune complexes were mM EDTA. further washed with 10 mM Tris (pH 8) and 100 mM NaCl before incubation Northern Blotting. Poly(A)ϩ RNA was isolated from breast epithelial cell in the same buffer with five units of calf intestinal phosphatase (Roche) for 45 ␮ min at 37°C. Precipitated proteins were released by boiling in sample buffer lines, and Northern analysis was performed as described (31). Briefly, 3 gof poly(A)ϩ RNA were fractionated on 1% agarose-formaldehyde gel, transferred and subjected to SDS-PAGE using 4–20% gradient gels (Novex, Carlsbad, ϩ CA). The proteins were blotted onto polyvinylidine difluoride membranes to Hybond N (Amersham Pharmacia Biotech), and hybridized with the (Millipore, Bedford, MA). After blocking with 20% horse serum (Hunter, following cDNA probes: (a) StarD10 cDNA, encompassing the entire 876-bp Jesmond, NSW, Australia) in PBS containing 0.1% Tween 20, filters were coding region; (b) ErbB2 cDNA (residues 1–228); or (c) glyceraldehyde-3- ϩ probed with specific antibodies. Proteins were visualized with peroxidase- phosphate dehydrogenase cDNA. A human multiple tissue poly(A) RNA coupled secondary antibody using the enhanced chemiluminescence detection Northern blot (Clontech Laboratories, Inc.) was hybridized with a StarD10 system (Amersham). Stripping of membranes was performed in SDS buffer cDNA probe. [62.5 mM Tris (pH 6.8), 2% SDS, and 100 mM ␤-mercaptoethanol] for 30 min Generation of StarD10-Specific Polyclonal Antibodies. Rabbits were at 60°C. Membranes were then reprobed with the indicated antibodies. immunized with StarD10 (264–277)-C264 peptide ESAVAESREERMG (Aus- Free Flow Electrophoresis (FFE). FFE was essentially performed as pep, Parkville, Victoria, Australia) coupled to keyhole limpet hemocyanin described in Hoffmann et al. (29) using the Octopus apparatus (Dr. Weber, (Sigma). Antibody was affinity purified with peptide that had been covalently Kirschheim GmbH, Germany). The isoelectric focusing running buffer was coupled to Sulfolink-coupling gel (Pierce, Rockford, IL). Elution was with 100 aqueous 0.2% w/v hydroxypropyl methyl cellulose and 0.2% w/v carrier mM glycine buffer (pH 2.7), and neutralized antibody-containing fractions ampholytes (Servalyte pH 3–10). Electrode solutions were 100 mM H3PO4 were pooled and dialyzed against PBS. (anode) and 50 mM NaOH (cathode); the counter flow solution (0.7 liter/min) Immunohistochemistry. Tissue arrays (Clinomics, Pittsfield, MA) were was 0.2% hydroxypropyl methyl cellulose containing 0.02 ML-arginine and blocked in PBS containing 5% goat serum (Hunter) and 0.05% Tween 20 and ϳ 0.02 ML-lysine. The sample was diluted to a concentration of 0.25 mg of then incubated with affinity-purified anti-StarD10 antibody, followed by in- protein/ml, and electrophoresis was performed at 4°C with a flow rate of 1.4 cubation with biotinylated antirabbit IgG and horseradish peroxidase-Strepta- ml/min and constant voltage of 1250 V. vidin (DAKO), before detection with diaminobenzidine (DAKO). Staining and Reverse-Phase High Pressure Liquid Chromatography (RP-HPLC). scoring with anti-ErbB2 antibody (DAKO) were performed using standard Analytical RP-HPLC was on a Smart System (Amersham Pharmacia) using a protocols (32). All antibody incubations were carried out in PBS containing Brownlee RP-300 (100 ϫ 2.1 mm, 300 Å,15␮m). A linear gradient from 0 to 100% B was applied at a flow rate of 100 ␮l/min, and 100-␮l fractions were 5% goat serum and 0.05% Tween 20. Slides were counterstained with hema- collected (A, 0.1% trifluoroacetic acid; B, 60% n-propyl alcohol). Preparative toxylin. ϫ 3 RP-HPLC was performed on a 1090 (Hewlett Packard, Palo Alto, CA) using Soft Agar Assays. HER14 cells (2 10 ) were plated in triplicate (six- a Brownlee RP 300 (100 ϫ 4.6 mm, 300 Å,15␮m). The flow rate was 500 well plates) in 2 ml of culture medium, supplemented with 0.35% DIFCO agar ␮l/min, and 500 ␮l fractions were collected. overlaying a 0.7% agar cushion. Cells were stained by the addition of 400 ␮l Protein Precipitation and Preparative SDS-PAGE. HPLC fractions con- of PBS containing 0.5 mg of nitroblue tetrazolium (Sigma) for 24 h before

taining the Mr 35,000 protein were pooled and precipitated with 0.5% deoxy- photographing using a Nikon SMZ-U microscope connected to an Axiocam cholate and 15% trichloroacetic acid (TCA) plus two volumes of acetone at digital camera (Zeiss, Jena, Switzerland). Photographs were analyzed using Ϫ20°C. After centrifugation at 16,000 ϫ g for 15 min, the protein pellet was ImageJ software (National Institutes of Health). 3539 STARD10 IN BREAST CANCER

RESULTS we devised a multistep biochemical purification strategy to purify the protein from human breast carcinoma SKBR3 cell lysates, which A Phosphospecific FKHR Antibody Cross-Reacts with a contain high levels of this protein. This strategy combined FFE, a Tumor-Specific M 35,000 Protein. The FOXO subgroup of fork- r continuous liquid-based, isoelectric-focusing method (29), RP-HPLC, head transcription factors consisting of FKHR, FKHRL1, and AFX and SDS-PAGE. By FFE, we identified two protein bands with was recently identified to be regulated by the PI3K pathway (17, 33). isoelectric-focusing points of pH 5.6 and pH 6.0 (Fig. 1B), which may These proteins drive the expression of cell cycle regulators, such as represent differential phosphorylation states of the protein. The M p27, and apoptotic proteins, including Bim and Fas ligand (17, 33). r 35,000 protein was also shown to elute in specific RP-HPLC fractions Phosphorylation on specific serine and threonine residues by protein (Fig. 1C). After preparative FFE purification of the M 35,000 protein kinase B/Akt leads to their inactivation and sequestration in the r from SKBR3 cytosolic protein, the pooled fractions containing the M cytoplasm by 14-3-3-scaffolding proteins, thereby preventing their r 35,000 protein were subjected to RP-HPLC and then SDS-PAGE negative effects on cell proliferation. (Fig. 1D). Bands migrating at M ϳ35,000 were excised, digested with To explore the potential inactivation of the FKHR transcription r trypsin in situ, and analyzed by tandem mass spectrometry. All 10 factor (FOXO1) by ErbB receptor signaling, mammary tumor lysates peptides identified from the indicated band in Fig. 1D were found to derived from mouse mammary tumor virus-neu transgenic mice were analyzed by Western blotting using an antiphosphopeptide antibody to match Q9Y365 (TrEMBL database; Fig. 2B), which corresponds to a FKHR phosphorylated on serine 256 (Fig. 1A). No signal correspond- human cDNA sequence conserved between the Caenorhabditis el- ing to FKHR was detected in tumor or adjacent mammary tissue. egans proteome and human expressed sequence tag nucleotide databases (28). Q9Y365 is the human homologue of mouse Pctp-l (phos- Interestingly, the antibody strongly cross-reacted with a Mr 35,000 protein that was present in tumor but not detectable in adjacent or phatidylcholine transfer protein-like), which was cloned from a testis normal mammary tissue (Fig. 1A). Immunoblotting with an anti-Neu/ library (34) and has been assigned the formal name StarD10. The ErbB2 antibody confirmed overexpression of the transgene in these predicted amino acid (aa) sequence did not contain a forkhead do- tumors. Further analysis of several breast epithelial cell lines revealed main, but a putative serine phosphorylation site resembling S256 in FKHR was identified (Fig. 2A). This provides a possible explanation high levels of the Mr 35,000 protein in the majority (6 of 11) of transformed epithelial lines (see Fig. 4B). Because no FOXO family for the observed cross-reactivity with the phospho-FKHR antibody. members corresponding to this molecular weight had been described, The Mr 35,000 Protein Shares Homology with Lipid-Binding Proteins. Sequence analysis of Q9Y365 predicted a molecular weight we speculated that the Mr 35,000 protein might represent a novel FKHR-related protein. This protein was likely to be phosphorylated of Mr 40,000. Given that no sequence coverage of the first predicted on a serine residue lying in a conserved sequence context, given the 71 aa was obtained, it seemed likely that translation initiated from specificity of the FKHR-specific antibody. methionine 69, giving rise to a protein of 291 residues with a theo- retical mass of M 33,000 (isoelectric point of 6.7; Fig. 2B). This Purification of the Mr 35,000 Protein. Because the phospho- r shorter version of Q9Y365 has recently been deposited into the FKHR antibody failed to immunoprecipitate the Mr 35,000 protein, Swiss-Prot database. To test whether this protein cross-reacted with the phospho-FKHR antibody, we cloned the corresponding cDNA (from SKBR3 cells) into the pEFrPGKpuro expression vector harbor-

ing an NH2-terminal Flag epitope tag. Indeed, transient transfection of this construct into 293T cells gave rise to a protein that was recognized by the phospho-FKHR antibody (Fig. 2C). The epitope tag accounted for the slight difference in size compared with the endog- enous protein in SKBR3 cells. Recognition of this protein proved to be phosphorylation dependent, because treatment of immunoprecipitated Flag-tagged StarD10 with alkaline phosphatase abolished cross- reactivity (Fig. 2D). This finding further indicates that serine 259 in StarD10 is likely to be phosphorylated. Database searches identified the presence of a START domain (35, 36) in StarD10 (aa 21–226), followed by a COOH-terminal tail of 65 aa. The START domain refers to a lipid-binding region first described in the steroidogenic acute regulatory protein StAR (37, 38). Although 15 START domain proteins have been identified in mammals thus far, lipid binding has only been established for a few members. To test the functionality of the START domain within StarD10, we made use of the intrinsic fluorescence of the protein. StarD10 contains multiple Fig. 1. Purification of a tumor-specific Mr 35,000 protein that cross-reacts with a phospho-FKHR antibody. In A, protein lysates derived from mammary tumors (T) and tyrosine and tryptophan residues that can be excited at 280 nm. A corresponding adjacent tissue (A) excised from B17, C13, and B14 mouse mammary change in the fluorescence emission spectrum in the presence of a tumor virus-Neu transgenic females or normal mammary gland tissue (N) were separated by SDS-PAGE and immunoblotted sequentially with phospho-FKHR-specific antibody binding partner is indicative of conformational changes within the (top panel), anti-ErbB2-, and antitubulin antibodies (middle and bottom panels). Lysate protein. In the absence of lipids, recombinant StarD10 protein purified from insulin-stimulated 293T cells transiently transfected with a FKHR expression construct was loaded as a positive control. In B and C, SKBR3 cytosolic protein lysate was from bacteria displayed a maxima fluorescence emission at 335 nm subjected to free flow electrophoresis (B) and reverse phase-high pressure liquid chro- (Fig. 2E). On addition of small unilamellar vesicles prepared with matography (C) as described. Fractions were separated by SDS-PAGE and analyzed by lipids extracted from porcine brain, a decrease in the intensity of immunoblotting using phospho-FKHR-specific antibody. In the case of the reverse phase-high pressure liquid chromatography samples, fractions (1 and 2, 3 and 4, etc.) were fluorescence was observed (Fig. 2E). This finding suggests that a pooled. D, Coomassie stain of precipitated fractions containing the Mr 35,000 protein conformational change has occurred in StarD10 upon interaction of separated by 10% SDS PAGE, after fractionation of SKBR3 cytosolic protein lysates by free flow electrophoresis and reverse phase-high pressure liquid chromatography. The the protein with membranes, indicating that StarD10 is capable of sequence shown in Fig. 2 corresponds to the band indicated by an arrow. binding lipids. 3540 STARD10 IN BREAST CANCER

Fig. 2. Cloning, expression, and lipid binding of the phosphoprotein StarD10. A, sequence alignment of Q9Y365 (indicated by an arrow in Fig. 1D) with FKHR using local similarity alignment. B, amino acid sequence of Q9Y365. Peptides identified by mass spectrometry are highlighted. Methi- onine 69 from which translation initiates is indicated. In C and D, 293T cells were transiently transfected with an expression construct encoding Flag-tagged StarD10. In C, lysates were separated by SDS-PAGE and immunoblotted using phospho- FKHR-specific antibody (top panel). SKBR3 protein lysate was loaded as a control. The membrane was reprobed with a Flag-specific antibody (bottom panel). In D, StarD10 was immunoprecipitated with anti-Flag antibody and incubated with or with- out calf intestinal phosphatase (CIP) for 45 min at 37°C before separation by SDS-PAGE. Immuno- blotting was performed with phospho-FKHR-specific antibody (top panel), followed by an anti-Flag monoclonal antibody (bottom panel). 293T whole- cell extract (WCE) containing transfected Flag- tagged StarD10 was loaded as a control. In E, the emission spectra of purified StarD10 protein (23 ␮g) in 20 mM HEPES (pH 7.2), 100 mM NaCl, and 5mM EDTA (1.8 ml) were recorded in the presence and absence of 20 nmol of brain lipid liposomes small unilamellar vesicles (SUVs)at37°C (excita- tion: 280 nm). Spectra were corrected for dilution and the absorbance of SUVs (a.u., arbitrary units).

Expression Analysis of StarD10. Northern analysis of StarD10 StarD10 expression, however, was also detected in cell lines that did mRNA from a variety of human organs revealed high levels of a not overexpress ErbB2. 1.4-kb transcript in liver, heart, skeletal muscle, and kidney, as also StarD10 Is Overexpressed in a Subset of Primary Human observed for the mouse homologue (Ref. 34; Fig. 3A). Although a Breast Cancers. To examine whether StarD10 was up-regulated in faint signal was detected in placenta, organs, including the brain, primary breast cancers, we performed immunohistochemistry on tis- colon, thymus, spleen, small intestine, lung, and peripheral blood sue arrays comprising 79 invasive breast cancers. Tumor specimens lymphocytes, lacked StarD10 mRNA. Examination of StarD10 ex- were scored as either low/negative or moderate/high for expression of pression in mouse tissues by immunoblotting confirmed high expres- StarD10, based on the intensity of staining with an affinity-purified sion in liver. Lower levels of StarD10 were detected in kidney, StarD10 antibody. A high proportion of tumors (28 of 79; 35%) salivary gland, testis, and colon (Fig. 3B). In the mammary gland, expression of StarD10 was found to be differentially regulated at different developmental stages. StarD10 was most abundant during pregnancy and lactation, with low expression evident in the virgin state (Fig. 3C). Coexpression of StarD10 and ErbB2 in Breast Carcinoma Lines. To generate an antibody specific for human StarD10, we immunized rabbits with a peptide encompassing aa 265–277 at the COOH terminus of this protein. Affinity-purified antibody specifically recognized the human StarD10 protein. Prominent expression of StarD10 was originally observed in tumors from mouse mammary tumor virus-neu transgenic mice and human breast carcinoma cell lines using a phospho-FKHR-specific antibody. To determine whether StarD10 was post-translationally modified in these carcinoma cell lines, we analyzed StarD10 by Western blotting using both phospho- FKHR- and StarD10-specific antisera (Fig. 4A). All breast carcinoma cell lines that cross-reacted with the phospho-FKHR antibody were found to express high levels of StarD10, indicating that the protein is overexpressed, and serine 295 is constitutively phosphorylated. In- triguingly, most cells lines that expressed ErbB2 also displayed high levels of StarD10 protein. We next sought to determine whether StarD10 up-regulation occurred at the RNA level and performed Northern analysis on a panel of breast epithelial cell lines (Fig. 4B). High levels of StarD10 RNA were found in the cell lines that express StarD10 protein, indicating Fig. 3. Tissue-specific expression of StarD10. In A, poly(A)ϩ RNA from different that increased expression occurs at the transcriptional level. Consist- human tissues (Clontech) was hybridized with a StarD10 cDNA probe. The larger transcript most likely corresponds to unspliced mRNA. In B and C, protein lysates ent with the results obtained by Western blotting, StarD10 mRNA was prepared from the indicated mouse tissues (B) and mouse mammary glands at different almost negligible in both human (184, HBL100) and mouse (SCp2, developmental stages (C) were separated by SDS-PAGE and immunoblotted with a StarD10-specific antibody (top panels). Equal loading was verified by reprobing with a Eph4, HC11) immortalized cell lines. Strikingly, all cell lines that tubulin-specific antibody (bottom panels). V, virgin; DP, day pregnancy; DL, day lacta- overexpressed ErbB2 mRNA were found to have high StarD10 levels. tion; DI, day involution. 3541 STARD10 IN BREAST CANCER

Fig. 4. StarD10 is overexpressed in breast carcinoma cell lines. In A, protein lysates prepared from the indicated cell lines were separated by SDS-PAGE and immunoblotted using ErbB2- (top panel), phospho-FKHR-, and StarD10-specific antibodies (middle panels). Equal loading was verified by reprobing with a tubulin-specific antibody (bottom panel). In B, poly(A)ϩ RNA (3 ␮g) from breast epithelial cell lines was hybridized with StarD10 (top panel) and ErbB2 (middle panel) cDNA probes. Equal loading was confirmed by hybridizing with a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA probe (bottom panel). yielded moderate to intense staining and included infiltrating ductal, transformation of NIH3T3 cells, it was found to enhance anchorage- lobular, and mixed carcinomas. Four infiltrating ductal breast carci- independent growth of HER14 cells in soft agar (Fig. 6B). Quantifi- nomas exhibiting intense staining of the nucleus and cytoplasm are cation of digital photographs using image analysis software revealed shown in Fig. 5, C–F. In contrast, negligible StarD10 expression was that StarD10 expression increased both the number and size of colo- observed in normal breast tissue (Fig. 5, A and B). Immunostaining nies (Fig. 6B). This was confirmed in multiple experiments and using with an anti-ErbB2 antibody revealed that of the eight tumors dis- independent pools of StarD10 HER14 transductants. These results playing strong membranous staining, four also expressed StarD10 at demonstrate that StarD10 cooperates with ErbB signaling in eliciting high levels. Two examples are depicted in Fig. 5, in which the anchorage-independent growth. StarD10-expressing tumors E and F correspond to the ErbB2-positive tumors G and H, respectively. StarD10 Cooperates with ErbB1 in Agar Assays. To investigate DISCUSSION the effect of overexpression of StarD10 on cellular transformation, we Aberrant expression of the ErbB2 receptor tyrosine kinase has been generated NIH3T3 and HER14 cells stably expressing StarD10 by implicated in the pathogenesis of a high proportion of sporadic breast retroviral infection. The HER14 line corresponds to NIH3T3 cells that cancers (2, 3). Moreover, targeted expression of ErbB2/Neu to the express ErbB1; these and parental NIH3T3 cells lack detectable mammary gland leads to tumor development in mice (9). Here, we StarD10 by Western blotting. Transduction of HER14 cells with report the isolation of the phosphoprotein StarD10 and its up-regula- StarD10 resulted in expression similar to that seen for StarD10 in tion in mouse mammary tumors, human breast carcinoma cell lines (9 breast carcinoma cell lines (Fig. 6A). Overexpression of StarD10 did of 14), and primary breast cancers (35%). A striking correlation not change the morphology of NIH3T3 or HER14 cells, nor was between ErbB2 status and StarD10 expression was found in Neu growth altered in either normal and limiting serum conditions (data tumors and human breast cancer cell lines. Moreover, StarD10 was not shown). Although StarD10 alone was not sufficient to cause demonstrated to cooperate with ErbB1 in cellular transformation

Fig. 5. StarD10 expression in primary human breast cancers. Immunohistochemistry with anti- StarD10 peptide antisera was performed on tissue arrays containing archival breast specimens. Neg- ligible expression was detected in normal breast tissue (A and B), whereas abundant StarD10 expression was evident in infiltrating ductal adeno- carcinomas (C–F). The specimens displayed in E and F were also positive for ErbB2/HER2, as shown in G and H, respectively.

3542 STARD10 IN BREAST CANCER

body, StarD10 appears to be constitutively phosphorylated on serine 259 in proliferating breast epithelial cells. This antibody recognizes FKHR phosphorylated on serine 256, a site targeted by PKB/Akt. However, the Ser259 site in StarD10 does not have the appropriate arginine determinants for protein kinase B/Akt, and treatment with the PI3K inhibitor LY294002 had no effect on phospho-FKHR antibody recognition of StarD10. The kinase responsible for phosphorylation of StarD10 at this site is not yet known. Functionally, the StarD10 S259A mutation has little effect on promoting anchorage-independent growth but may be important for regulating protein turnover, because the mutant protein was expressed at a lower level (data not shown). Of the 15 mammalian START domain-containing proteins, which have been assigned the formal names StarD1–15, Pctp-l/StarD10 displays the highest homology with the Pctp subfamily, consisting of Pctp/StarD2, GTT1/StarD7, and GPBP/StarD11 (41). The START domain protein MLN64 is coamplified with ErbB2 in human tumors (42). Similar to StAR, the founding member of this family, MLN64 is a cholesterol transfer protein that is thought to deliver cholesterol to the inner mitochondrial membrane, where it serves as a metabolic precursor for steroid hormones (43–45). The up-regulation of MLN64 has therefore been proposed to contribute to intratumoral steroidogen- Fig. 6. StarD10 cooperates with ErbB1/epidermal growth factor receptor to promote esis. Pctp exclusively binds PC (46). The quenching of the intrinsic growth in soft agar. In A, protein lysates derived from HER14 cells transduced with fluorescence of StarD10 by brain lipid liposomes (Fig. 2E) indicates pBabePuro (control) or pBabeStarD10 recombinant retroviruses were separated by SDS- PAGE and immunoblotted with StarD10-specific antibody. In B, 2 ϫ 103 HER14 cells that StarD10 also interacts with lipids. Preliminary data suggest that infected with either control Puro or StarD10-expressing virus were plated in 0.35% top StarD10 can bind phosphatidylcholine,6 but its specificity for other agar to assay anchorage-independent growth. Photographs of triplicate wells were taken lipids remains to be defined. after staining with nitroblue tetrazolium, and quantification of colony numbers was performed using ImageJ software. The number of HER14 Puro colonies was set to 100%, Both StarD10 and Pctp are particularly abundant in the liver, where and error bars represent SEM. they may be involved in export of lipids into bile (47, 48). The presence of high levels of StarD10 in the lactating mammary gland assays, in which an increase in both the number and size of colonies suggests a potential physiological role in the functional differentiation was evident. In primary carcinomas, 50% (4 of 8) of ErbB2-positive of this organ. Within the cell, lipids are used by phospholipases to tumors expressed high levels of StarD10. These data indicate a strong generate second messengers, such as phosphatidic acid and diacyl- correlation between StarD10 and ErbB2 overexpression, suggesting a glyerol (49). StarD10 may play a role in replenishing membranes with selective growth advantage for tumors expressing both proteins. De- specific lipids metabolized by phospholipases, and deregulation of spite coexpression of StarD10 and ErbB2 in breast cancer cell lines this transport may lead to aberrant lipid signaling, thereby contribut- and a subset of breast tumors, StarD10 does not appear to be a direct ing to cellular transformation. The phosphatidylinositol transfer pro- transcriptional target of the receptor. In NIH3T3 cells that express tein has been demonstrated to influence EGF signaling, whereby constitutively active Neu (NeuT), no increase in the level of StarD10 EGF-induced phosphorylation of phosphatidylinositol 3-phosphate protein was observed (data not shown). The molecular basis of the and subsequent hydrolysis by phospholipase C ␥ was shown to require apparent coexpression of StarD10 and ErbB2 remains to be estab- the presence of phosphatidylinositol transfer protein (50). It is possi- lished but may involve regulation by a common transcription factor. ble that StarD10 fulfills a similar function as a cofactor for receptor- Interestingly, StarD10 localizes to human chromosome 11q13, in mediated lipid signaling. In HER14 cells, the expression of StarD10 proximity to the amplicon comprising the cyclin D1 and EMS genes, did not increase EGF-induced PI3K activation (data not shown), but which are frequently overexpressed in breast cancer (39). However, it may affect other signaling pathways. Alternatively, the ability of the mechanism underlying StarD10 overexpression in breast cancer StarD10 to stimulate anchorage-independent growth and cooperate appears to reflect increased transcription rather than gene amplifica- with ErbB receptors may be independent of its lipid-binding function. tion, because Southern analysis of DNA from normal and human StarD10 has a 65 aa COOH-terminal extension that is not present in breast cancer cell lines revealed no alterations (data not shown). Pctp and may confer an additional function. Elucidation of the func- Overexpression of StarD10 may occur in additional tumor types, tional domains within StarD10 will provide insight into the mecha- such as colon carcinoma. Screening of a cDNA expression library nism by which this protein acts. derived from a human colon tumor patient with autologous serum identified autoantibodies against an expressed sequence tag corresponding to StarD10 (40). This finding indicates that StarD10 may be ACKNOWLEDGMENTS a target for antitumoral immune response. Furthermore, we have evidence that StarD10 is up-regulated in a significant proportion of We thank Yoshitake Nishimune for the generous gift of Pctp-like antibody, human colon carcinoma cell lines (data not shown), indicating a Frosa Katsis for peptide conjugation, Rob Moritz and Ed Nice for advice on potential role in the development of this cancer. ErbB1 is frequently RP-HPLC, Joan Heath for colon carcinoma samples, Roger Daly for providing overexpressed in colorectal cancer. However, there is no apparent HER14 cells, and Stephen Jane for supplying a Clontech poly(A)ϩ blot. We correlation between StarD10 expression and that of either ErbB1 or also thank Belinda Duscio for technical assistance, Samantha Oakes for help ErbB2 in these cell lines. StarD10 may therefore synergize with other with image analysis, and Tim Beissbarth and Sergio Wittlin for discussions. oncogenes in colon cell neoplasia. On the basis of our data using the phospho-FKHR-specific anti- 6 Unpublished data. 3543 STARD10 IN BREAST CANCER

REFERENCES 25. Ottenhoff-Kalff AE, Rijksen G, van Beurden EA, Hennipman A, Michels AA, Staal GE. Characterization of protein tyrosine kinases from human breast cancer: involve- 1. Slamon DJ, Clark GM, Wong SG, Levin WJ, Ullrich A, McGuire WL. Human breast ment of the c-src oncogene product. Cancer Res 1992;52:4773–8. cancer: correlation of relapse and survival with amplification of the HER-2/neu 26. Verbeek BS, Vroom TM, Adriaansen-Slot SS, et al. c-Src protein expression is oncogene. Science 1987;235:177–82. increased in human breast cancer. An immunohistochemical and biochemical analy- 2. Salomon DS, Brandt R, Ciardiello F, Normanno N. Epidermal growth factor-related sis. J Pathol 1996;180:383–8. peptides and their receptors in human malignancies. Crit Rev Oncol Hematol 1995; 27. Maa MC, Leu TH, McCarley DJ, Schatzman RC, Parsons SJ. Potentiation of epider- 19:183–232. mal growth factor receptor-mediated oncogenesis by c-Src: implications for the 3. Mendelsohn J, Baselga J. The EGF receptor family as targets for cancer therapy. etiology of multiple human cancers. Proc Natl Acad Sci USA 1995;92:6981–5. Oncogene 2000;19:6550–65. 28. Lai CH, Chou CY, Ch’ang LY, Liu CS, Lin W. Identification of novel human genes 4. Olayioye MA, Neve RM, Lane HA, Hynes NE. The ErbB signaling network: receptor evolutionarily conserved in Caenorhabditis elegans by comparative proteomics. Ge- heterodimerization in development and cancer. EMBO J 2000;19:3159–67. nome Res 2000;10:703–13. 5. Yarden Y, Sliwkowski MX. Untangling the ErbB signalling network. Nat Rev Mol. 29. Hoffmann P, Ji H, Moritz RL, et al. Continuous free-flow electrophoresis separation Cell Biol 2001;2:127–37. of cytosolic proteins from the human colon carcinoma cell line LIM 1215: a non 6. Troyer KL, Lee DC. Regulation of mouse mammary gland development and tumor- two-dimensional gel electrophoresis-based proteome analysis strategy. Proteomics igenesis by the ERBB signaling network. J Mammary Gland Biol Neoplasia 2001; 2001;1:807–18. 6:7–21. 30. Perkins DN, Pappin DJ, Creasy DM, Cottrell JS. Probability-based protein identifi- 7. Garrett TP, McKern NM, Lou M, et al. The crystal structure of a truncated ErbB2 cation by searching sequence databases using mass spectrometry data. Electrophore- ectodomain reveals an active conformation, poised to interact with other ErbB sis 2003;20:3551–67. receptors. Mol Cell 2003;11:495–505. 31. Visvader J, Begley CG, Adams JM. Differential expression of the LYL, SCL and E2A 8. Stern DF. ErbBs in mammary development. Exp Cell Res 2003;284:89–98. helix-loop-helix genes within the hemopoietic system. Oncogene 1991;6:187–94. 9. Guy CT, Webster MA, Schaller M, Parsons TJ, Cardiff RD, Muller WJ. Expression 32. Provenzano E, Hopper JL, Giles GG, Marr G, Venter DJ, Armes JE. Biological of the neu protooncogene in the mammary epithelium of transgenic mice induces markers that predict clinical recurrence in ductal carcinoma in situ of the breast. Eur metastatic disease. Proc Natl Acad Sci USA 1992;89:10578–82. J Cancer 2003;39:622–30. 10. Siegel PM, Muller WJ. Mutations affecting conserved cysteine residues within the 33. Brazil DP, Park J, Hemmings BA. PKB binding proteins. Getting in on the Akt. Cell extracellular domain of Neu promote receptor dimerization and activation. Proc Natl 2002;111:293–303. Acad Sci USA 1996;93:8878–83. 34. Yamanaka M, Koga M, Tanaka H, et al. Molecular cloning and characterization of 11. Siegel PM, Ryan ED, Cardiff RD, Muller WJ. Elevated expression of activated forms phosphatidylcholine transfer protein-like protein gene expressed in murine haploid of Neu/ErbB-2 and ErbB-3 are involved in the induction of mammary tumors in germ cells. Biol Reprod 2000;62:1694–701. transgenic mice: implications for human breast cancer. EMBO J 1999;18:2149–64. 35. Ponting CP, Aravind L. START: a lipid-binding domain in StAR, HD-ZIP and 12. Holbro T, Beerli RR, Maurer F, Koziczak M, Barbas CF III, Hynes NE. The signaling proteins. Trends Biochem Sci 1999;24:130–2. ErbB2/ErbB3 heterodimer functions as an oncogenic unit: ErbB2 requires ErbB3 to 36. Schultz J, Copley RR, Doerks T, Ponting CP, Bork P. SMART: a web-based tool for drive breast tumor cell proliferation Proc Natl Acad Sci USA 2003;100:8933–8. the study of genetically mobile domains. Nucleic Acids Res 2000;28:231–4. 13. Alimandi M, Romano A, Curia MC, et al. Cooperative signaling of ErbB3 and ErbB2 37. Christenson LK, Strauss JF III. Steroidogenic acute regulatory protein: an update on in neoplastic transformation and human mammary carcinomas. Oncogene 1995;10: its regulation and mechanism of action. Arch Med Res 2001;32:576–86. 1813–21. 38. Stocco DM. StAR protein and the regulation of steroid hormone biosynthesis. Annu 14. Pinkas-Kramarski R, Soussan L, Waterman H, et al. Diversification of Neu differ- Rev Physiol 2001;63:193–213. entiation factor and epidermal growth factor signaling by combinatorial receptor 39. Ormandy CJ, Musgrove EA, Hui R, Daly RJ, Sutherland RL. Cyclin D1, EMS1 and interactions. EMBO J 1996;15:2452–67. 11q13 amplification in breast cancer. Breast Cancer Res Treat 2003;78:323–35. 15. Fedi P, Pierce JH, di Fiore PP, Kraus MH. Efficient coupling with phosphatidyli- 40. Scanlan MJ, Chen YT, Williamson B, et al. Characterization of human colon cancer nositol 3-kinase, but not phospholipase C ␥ or GTPase-activating protein, distin- antigens recognized by autologous antibodies. Int J Cancer 1998;76:652–8. guishes ErbB-3 signaling from that of other ErbB/EGFR family members. Mol Cell 41. Soccio RE, Breslow JL. StAR-related lipid transfer (START) proteins: mediators of Biol 1994;14:492–500. intracellular lipid metabolism. J Biol Chem 2003;278:22183–6. 16. Prigent SA, Gullick WJ. Identification of c-erbB-3 binding sites for phosphatidyli- 42. Moog-Lutz C, Tomasetto C, Regnier CH, et al. MLN64 exhibits homology with the nositol 3Ј-kinase and SHC using an EGF receptor/c-erbB-3 chimera. EMBO J steroidogenic acute regulatory protein (STAR) and is over-expressed in human breast 1994;13:2831–41. carcinomas. Int J Cancer 1997;71:183–91. 17. Burgering BM, Kops GJ. Cell cycle and death control: long live Forkheads. Trends 43. Watari H, Arakane F, Moog-Lutz C, et al. MLN64 contains a domain with homology Biochem Sci 2002;27:352–60. to the steroidogenic acute regulatory protein (StAR) that stimulates steroidogenesis. 18. Ricci A, Lanfrancone L, Chiari R, et al. Analysis of protein-protein interactions Proc Natl Acad Sci USA 1997;94:8462–7. involved in the activation of the Shc/Grb-2 pathway by the ErbB-2 kinase. Oncogene 44. Tsujishita Y, Hurley JH. Structure and lipid transport mechanism of a StAR-related 1995;11:1519–29. domain. Nat Struct Biol 2000;7:408–14. 19. Dankort DL, Wang Z, Blackmore V, Moran MF, Muller WJ. Distinct tyrosine 45. Zhang M, Liu P, Dwyer NK, et al. MLN64 mediates mobilization of lysosomal autophosphorylation sites negatively and positively modulate neu-mediated transfor- cholesterol to steroidogenic mitochondria. J Biol Chem 2002;277:33300–10. mation. Mol Cell Biol 1997;17:5410–25. 46. Kamp HH, Wirtz WA, Baer PR, et al. Specificity of the phosphatidylcholine ex- 20. Avruch J, Zhang XF, Kyriakis JM. Raf meets Ras: completing the framework of a change protein from bovine liver. Biochemistry 1977;16:1310–6. signal transduction pathway. Trends Biochem Sci 1994;19:279–83. 47. Cohen DE, Leonard MR, Carey MC. In vitro evidence that phospholipid secretion 21. Marais R, Light Y, Paterson HF, Marshall CJ. Ras recruits Raf-1 to the plasma into bile may be coordinated intracellularly by the combined actions of bile salts and membrane for activation by tyrosine phosphorylation. EMBO J 1995;14:3136–45. the specific phosphatidylcholine transfer protein of liver. Biochemistry 1994;33: 22. Sinn E, Muller W, Pattengale P, Tepler I, Wallace R, Leder P. Coexpression of 9975–80. MMTV/v-Ha-ras and MMTV/c-myc genes in transgenic mice: synergistic action of 48. LaMorte WW, Booker ML, Kay S. Determinants of the selection of phosphatidyl- oncogenes in vivo. Cell 1987;49:465–75. choline molecular species for secretion into bile in the rat. Hepatology 1998;28: 23. Cheng AM, Saxton TM, Sakai R, et al. Mammalian Grb2 regulates multiple steps in 631–7. embryonic development and malignant transformation. Cell 1998;95:793–803. 49. Nishizuka Y. Intracellular signaling by hydrolysis of phospholipids and activation of 24. Muthuswamy SK, Siegel PM, Dankort DL, Webster MA, Muller WJ. Mammary protein kinase C. Science, 1992;258:607–14. tumors expressing the neu proto-oncogene possess elevated c-Src tyrosine kinase 50. Kauffmann-Zeh A, Thomas GM, Ball A, et al. Requirement for phosphatidylinositol activity. Mol Cell Biol 1994;14:735–43. transfer protein in epidermal growth factor signaling. Science 1995;268:1188–90.

3544