Global Characterization of Signalling Networks Associated with Tamoxifen Resistance in Breast Cancer Brigid C

Global characterization of signalling networks associated with tamoxifen resistance in breast cancer Brigid C. Browne1, Falko Hochgrafe€ 2, Jianmin Wu1,3, Ewan K. A. Millar1,4,5,6, Jane Barraclough1, Andrew Stone1, Rachael A. McCloy1, Christine S. Lee1, Caroline Roberts1, Naveid A. Ali1, Alice Boulghourjian1, Fabian Schmich7,8, Rune Linding9, Lynn Farrow10, Julia M. W. Gee10, Robert I. Nicholson10, Sandra A. O’Toole1,3,11,12, Robert L. Sutherland1,3,†, Elizabeth A. Musgrove1,3, Alison J. Butt1,3,*,‡ and Roger J. Daly1,3,13,‡

1 Cancer Research Program, The Kinghorn Cancer Centre, Garvan Institute of Medical Research, Sydney, New South Wales, Australia 2 Competence Center – Functional Genomics, Junior Research Group Pathoproteomics, University of Greifswald, Germany 3 St Vincent’s Clinical School, Faculty of Medicine, University of New South Wales, Darlinghurst, New South Wales, Australia 4 School of Medical Sciences, Faculty of Medicine, University of New South Wales, Darlinghurst, New South Wales, Australia 5 Department of Anatomical Pathology, South Eastern Area Laboratory Service, St George Hospital, Kogarah, New South Wales, Australia 6 School of Medicine and Health Sciences, University of Western Sydney, Campbelltown, New South Wales, Australia 7 Department of Biosystems Science and Engineering, ETH Zurich, Basel, Switzerland 8 Swiss Institute of Bioinformatics, Basel, Switzerland 9 Cellular Signal Integration Group, Center for Biological Sequence Analysis, Department of Systems Biology, Technical University of Denmark, Lyngby, Denmark 10 Cardiff School of Pharmacy & Pharmaceutical Sciences, Cardiff University, UK 11 Department of Tissue Pathology and Diagnostic Oncology, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia 12 Sydney Medical School, University of Sydney, Westmead,New South Wales, Australia 13 Department of Biochemistry and Molecular Biology, School of Biomedical Sciences, Monash University, Melbourne,Victoria, Australia

Keywords Acquired resistance to the anti-estrogen tamoxifen remains a significant endocrine resistance; MARCKS; MCF7; challenge in breast cancer management. In this study, we used an integrative phosphoproteomics; Yes kinase approach to characterize global protein expression and tyrosine phosphorylation events in tamoxifen-resistant MCF7 breast cancer cells (TamR) Correspondence compared with parental controls. Quantitative mass spectrometry and compu- R. J. Daly, Department of Biochemistry and tational approaches were combined to identify perturbed signalling networks, Molecular Biology, School of Biomedical and candidate regulatory proteins were functionally interrogated by siRNA- Sciences, Level 1, Building 77, Monash mediated knockdown. Network analysis revealed that cellular metabolism was University, Melbourne, Victoria 3800, perturbed in TamR cells, together with pathways enriched for proteins associ- Australia ated with growth factor, cell–cell and cell matrix-initiated signalling. Consistent Fax: +61 3 990 29500 with known roles for Ras/MAPK and PI3-kinase signalling in tamoxifen resis- Tel: +61 3 990 29301 tance, tyrosine-phosphorylated MAPK1, SHC1 and PIK3R2 were elevated in E-mail: [email protected] TamR cells. Phosphorylation of the tyrosine kinase Yes and expression of the actin-binding protein myristoylated alanine-rich C-kinase substrate (MARCKS) *Present address were increased two- and eightfold in TamR cells respectively, and these pro- Research Investment, National Breast teins were selected for further analysis. Knockdown of either protein in TamR Cancer Foundation, 50 Pitt Street, Sydney, cells had no effect on anti-estrogen sensitivity, but significantly decreased cell New South Wales, Australia motility. MARCKS expression was significantly higher in breast cancer cell lines than normal mammary epithelial cells and in ER-negative versus ER-posi- †Deceased tive breast cancer cell lines. In primary breast cancers, cytoplasmic MARCKS ‡These authors contributed equally to this staining was significantly higher in basal-like and HER2 cancers than in lumi- work nal cancers, and was independently predictive of poor survival in multivariate analyses of the whole cohort (P < 0.0001) and in ER-positive patients (Received 18 April 2013, revised 27 June (P = 0.0005). These findings provide network-level insights into the molecular 2013, accepted 17 July 2013) alterations associated with the tamoxifen-resistant phenotype, and identify doi:10.1111/febs.12441 MARCKS as a potential biomarker of therapeutic responsiveness that may assist in stratification of patients for optimal therapy.

Abbreviations EGFR, epidermal growth factor receptor; ER, estrogen receptor; MARCKS, myristoylated alanine-rich C-kinase substrate; OHT, 4-hydroxytamoxifen; pY, phosphotyrosine; SFK, Src family kinase; SILAC, stable isotopic labelling using amino acids in culture; TamR, tamoxifen-resistant MCF7 cell line.

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5237 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al.

Introduction The anti-estrogen tamoxifen has been the most widely When combined with appropriate enrichment tech- used endocrine treatment for breast cancer for more niques, mass spectrometry may also be used to charac- than 30 years, and has significantly improved survival terize changes in particular sub-proteomes, such as the for patients with estrogen receptor (ER)-positive dis- fraction of cellular proteins that are tyrosine-phosphor- ease [1]. However, approximately one-third of patients ylated [29], providing global insights into perturbations treated with tamoxifen for 5 years develop recurrent in particular signalling and regulatory processes. In this disease within 15 years [2]. Responses to second-line study, we have used an integrated approach that com- therapeutic strategies, including sequential delivery of bines quantitative proteomics and phosphoproteomics alternative endocrine therapies, therapies targeted to with bioinformatics, functional interrogation and anal- growth factor receptors, and, in some instances, che- ysis of breast cancer cohorts in order to characterize motherapy, are often short-lived, and disease progres- changes in cellular signalling networks associated with sion is inevitable [3–6]. Further deciphering the acquisition of tamoxifen resistance. Our work high- molecular basis of tamoxifen resistance is therefore crit- lights key regulatory pathways associated with this phe- ical in order to identify robust biomarkers of response, notype, as well as the complexity of the associated as well as targets for potential therapeutic intervention. network perturbations, and identifies myristoylated To this end, early candidate-based studies of tamoxi- alanine-rich C-kinase substrate (MARCKS) as a candi- fen resistance revealed roles for receptor tyrosine kin- date prognostic marker in ER-positive breast cancer. ases such as epidermal growth factor receptor (EGFR), human epidermal growth factor receptor 2 Results (HER2/ERBB2) and insulin-like growth factor 1 receptor [7–9], as well as numerous intracellular signal- Workflow for SILAC and MS identification of ling mediators [10–14]. Genome-wide gene expression peptides analyses performed on both experimental models of endocrine resistance and ER-positive patient samples The tamoxifen-resistant (TamR) cell line was devel- have also been used to identify potential mechanisms oped by long-term culture of MCF7 cells in the pres- of resistance and generate gene signatures that predict ence of 4-hydroxytamoxifen (OHT) [9]. Figure 1 tamoxifen response [15–25]. We have previously describes the workflow for SILAC (stable isotopic defined functionally distinct gene signatures represent- labelling using amino acids in culture) and MS analy- ing cell proliferation, apoptosis and cell growth, each ses of MCF7 and TamR cells. Briefly, equal propor- of which was predictive of response in a cohort of tions of TamR and MCF7 cells labelled with heavy tamoxifen-treated patients [19]. Gene signatures, such and light isotopes, respectively, were mixed, and iso- as those included in the commercially available Onco- lated proteins were separated by SDS/PAGE. This was typeDx (Genomic Health, Redwood City, CA, USA) followed by in-gel digestion with trypsin, extraction of [20] and Prosigna (Nanostring Technologies, Seattle, peptides and MS analysis (Fig. 1A). In parallel, equal WA, USA) [25] gene assays predict residual risk of dis- proportions of differentially SILAC-labelled MCF7 tant relapse in specific subsets of ER-positive patients, and TamR cells were mixed and protein lysates were and thus guide treatment choices. However, it is well digested overnight. Phosphotyrosine (pY) peptides understood that cellular mRNA levels do not always were eluted after sequential immunoprecipitation with reflect the level of the encoded protein nor the vast pY100 and pY20 antibodies, and MS analyses were array of post-translational modifications that individ- performed (Fig. 1B). Each assay was performed in ual proteins undergo during regulation of cellular duplicate, with reversed labelling of each cell line. In processes. total, 2059 proteins were identified by total proteomics MS-based proteomic technology is being increasingly analysis, and 116 pY sites were identified by phospho- used to study and compare the proteomes of in vitro proteomic analysis (Tables S1B and S2B). and in vivo models of cancer as well as patient tumours, and has augmented knowledge obtained from Protein expression changes associated with gene expression profiling. A handful of studies have tamoxifen resistance used this tool to examine tamoxifen resistance, and have revealed significant differences in the expression A total of 186 proteins were differentially expressed in of proteins associated with key biological processes, TamR relative to MCF7 cells, with 86 proteins up-regu- such as cell proliferation, survival and motility [26–28]. lated and 100 down-regulated by 1.5-fold or more. The

5238 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

SDS/PAGE A Total proteome Peptides TamR Tryptic digest 50 ug Heavy Lyse IP pY100 pY20 MS- Mix antibodies identification 1:1 Light 40 mg Elute MCF7 pY-peptides KLpYQVPT (2 replicate assays, wash non-pY inversed labelling) B pY peptides

Fig. 1. Workﬂow for SILAC and MS identiﬁcation of peptides. (A) Equal amounts of lysate from TamR and MCF7 cells labelled with heavy and light isotopes, respectively, were mixed and proteins were separated by SDS/PAGE, followed by in-gel digestion with trypsin, extraction of peptides and MS analysis. (B) Phosphotyrosine peptides were immunoprecipitated sequentially with pY100 and pY20 antibodies. Following peptide elution, MS analyses were performed. Each assay was performed in duplicate, with reversed labelling of each cell line.

186 differentially expressed proteins are listed in Table Perturbations in tyrosine phosphorylation S1, and the 12 proteins exhibiting the greatest up- or associated with tamoxifen resistance down-regulation are presented in Table 1. Full names of abbreviated proteins and genes mentioned in this Fifty pY sites, representing 43 unique proteins, dif- paper are presented in Tables S1 and S2, and further fered in abundance by 1.2-fold or more between details are available at http://www.genecards.org. MCF7 and TamR cells in two replicate assays using

Table 1. Twelve proteins with the highest levels of upregulation and downregulation in TamR cells relative to MCF7 cells. The complete list of 186 differentially expressed proteins is in Table S1. SD, standard deviation.

Mean ratio Accession Gene Protein name TAMR/MCF7 SD

P25815 S100P S100 calcium binding protein P 31.17 21.91 P52895 AKR1C2 Aldo-keto reductase family 1, member C 18.13 1.24 P05026 ATP1B1 ATPase, Na+/K+ transporting, beta 1 polypeptide 13.96 14.51 P29966 MARCKS Myristoylated alanine-rich protein kinase C substrate 7.87 0.76 P15559 NQO1 NAD(P)H dehydrogenase, quinone 1 3.94 0.21 O14975 SLC27A2 Solute carrier family 27 (fatty acid transporter), member 2 3.73 1.53 P29762 CRABP1 Cellular retinoic acid binding protein 1 2.79 0.07 Q16881 TXNRD1 Thioredoxin reductase 1 2.77 0.65 P30520 ADSS Adenylosuccinate synthase 2.66 0.89 P06737 PYGL Phosphorylase, glycogen, liver 2.58 0.06 P60903 S100A10 S100 calcium binding protein A10 2.53 0.34 P55957 BID BH3 interacting domain death agonist 2.49 0.13 Q9BTT0 ANP32E Acidic (leucine-rich) nuclear phosphoprotein 32 family, member E 0.29 0.07 P09467 FBP1 Fructose-1,6-bisphosphatase 1 0.28 0.01 Q9BS40 LXN Latexin 0.27 0.00 P22307 SCP2 Sterol carrier protein 2 0.26 0.01 Q9HCY8 S100A14 S100 calcium binding protein A14 0.26 0.05 O43805 SSNA1 Sjogren syndrome nuclear autoantigen 1 0.25 0.02 Q15847 C10orf116 Chromosome 10 open reading frame 116 0.22 0.02 Q8TB36 GDAP1 Ganglioside-induced differentiation-associated protein 1 0.20 0.07 O95817 BAG3 BCL2-associated athanogene 3 0.20 0.09 P06703 S100A6 S100 calcium binding protein A6 0.20 0.06 Q9UJY1 HSPB8 Heat shock 22 kDa protein 8 0.17 0.03 P29373 CRABP2 Cellular retinoic acid binding protein 2 0.14 0.02

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5239 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al.

Table 2. Eighteen tyrosine sites identiﬁed with altered levels in both pY100 and pY20 phosphoproteomic analyses. The complete list of all differentially phosphorylated pY sites is in Table S2. SD, standard deviation.

Mean ratio Accession Gene Protein name Position TamR/MCF7 SD

Q9UPX8 SHAN2 Isoform E of SH3 and multiple ankyrin repeat domains protein 2 Y_989 3.92 1.37 P11413 G6PD Isoform 3 of Glucose-6-phosphate 1-dehydrogenase Y_431 2.78 0.05 P07947 YES1 Proto-oncogene tyrosine-protein kinase Yes Y_222 2.07 0.06 P07947 YES1 Proto-oncogene tyrosine-protein kinase Yes Y_426 1.98 0.08 P28482 MAPK1 Mitogen-activated protein kinase 1 (MK01) Y_187 1.68 0.12 O15357 SHIP2 Phosphatidylinositol-3,4,5-trisphosphate 5-phosphatase 2 Y_886 1.49 0.33 Q9NRY4 GRLF1 Glucocorticoid receptor DNA-binding factor 1 Y_1105 1.45 0.06 P29353 SHC1 SHC-transforming protein 1 Y_427 1.41 0.10 P46109 CRKL Crk-like protein Y_207 1.37 0.08 Q63HR2 TENC1 Isoform 4 of Tensin-like C1 domain-containing phosphatase Y_493 1.32 0.08 P27361 MAPK3 Mitogen-activated protein kinase 3 (MK03) Y_204 0.61 0.03 O00401 WASL Neural Wiskott-Aldrich syndrome protein Y_256 0.58 0.03 Q9Y2H5 PKHA6 Pleckstrin homology domain-containing family A member 6 Y_492 0.58 0.00 Q8TEW0 PARD3 Partitioning defective 3 homolog Y_1080 0.47 0.03 P35568 IRS1 Insulin receptor substrate 1 Y_612 0.43 0.03 P54764 EPHA4 Ephrin type-A receptor 4 Y_779 0.32 0.01 P54760 EPHB4 Ephrin type-B receptor 4 Y_574 0.26 0.02 O60716 CTNND1 Catenin delta-1 Y_904 0.20 0.02

either of the pY antibodies (Table S2A). A consider- ably higher number of pY sites were down-regulated

(36 sites) than were up-regulated (14 sites) in TamR MCF7 TamR MCF7 TamR relative to MCF7 cells. Eighteen pY sites, representing STAT3 MARCKS 17 proteins, differed based on two replicates of both pY705 the pY100 and pY20 analyses (Table 2). MARCKS STAT3 pS152/6 Validation of altered protein expression and phosphorylation MAPK3 BID MAPK1 Western blotting was performed to confirm differential pY204 phosphorylation and/or expression of a number of the Ezrin identified proteins and phosphoproteins (Fig. 2). MAPK3 Increased expression of MARCKS and BID and MAPK1 reduced Ezrin levels were confirmed in TamR cells. Yes IRS1 These cells also contained elevated levels of MARCKS pY416 pY612 phosphorylated on S152/6. Phosphorylation of Y705 Yes total on STAT3 and Y612 on IRS1 were increased and IRS1 decreased, respectively, but the total expression level of either protein was unaltered, highlighting the ability Src total of the phosphoproteomics approach to identify Actin changes at the post-translational level. Increased phos- Actin phorylation of Yes at Y222 was demonstrated by MS analysis (Table 2). However, while increased phosphorylation at Y426 was also indicated (Table 2), the Fig. 2. Validation of MS data by western blot analysis. Lysates derived from MCF7 and TamR cells were western blotted with the assignment of this active-site phosphopeptide to Yes is indicated antibodies. The arrow in the left panel indicates not unequivocal as the peptide sequence is shared by phosphorylated Yes at 62 kDa. The relative mobility of other Src family kinases (SFKs). In order to confirm phosphorylated Yes was confirmed by siRNA-mediated knockdown the differential phosphorylation of Yes at this site, we (see Fig. 5).

5240 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance exploited the slightly lower mobility of Yes (61 kDa) protein–protein interaction analysis using the 186 pro- upon SDS/PAGE versus other SFKs. Blotting using a teins identified in the total proteomic analysis and 43 phospho-specific antibody against the active site of phosphoproteins identified in the phosphoproteomic SFKs confirmed increased relative phosphorylation of analysis. STAT3 was included in the phosphoprotein Yes at this site (Fig. 2). Interestingly, differential regu- analyses, as, although increased STAT3 phosphoryla- lation of p42 MAPK (MAPK1/MK01) and p44 tion was detected in only one replicate of each of the MAPK (MAPK3/MK03) was indicated by our phos- pY100 and pY20 analyses, this result was validated by phoproteomic profiling (Table 2). Western blotting western blotting. We first undertook pathway enrich- using a phospho-specific antibody that recognizes the ment analysis (Table 3). For differentially expressed active site of both MAPKs revealed that, for MAPK1, proteins (exhibiting either up- or down-regulation), the a change in activation level rather than expression was only pathway exhibiting significant enrichment was responsible for the observed increase in phosphoryla- ‘metabolism’, which reflected altered expression of a tion of this site in TamR cells, while for MAPK3, variety of metabolic enzymes involved in biosynthesis a change in expression level was the major contribut- (HMBS, NAMPT and UGDH), carbohydrate and ing factor (Fig. 2). fatty acid metabolism (PYGL and ACOX1), and gluco- neogenesis (FBP1), as well as MARCKS, which is involved in insulin signalling [30]. In contrast, when Pathway and network analyses changes in the tyrosine phosphoproteome were consid- In order to obtain functional insights into proteins ered, several pathways or processes showed a signifi- exhibiting differential expression or tyrosine phosphor- cant P value, or a strong trend towards significance. ylation between tamoxifen-sensitive and -resistant cells, These included ‘adherens junction’, ‘aldosterone- we first undertook pathway enrichment analysis and regulated sodium reabsorption’, ‘bacterial invasion of

Table 3. Pathway enrichment analysis of TamR cells showing pathways identiﬁed from differentially expressed and differentially phosphorylated proteins.

No. Pathway Database Proteins P Corrected P Genes

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5241 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al. epithelial cells’, ‘chemokine signalling pathway’, ‘neuro- reduced Yes protein expression in both MCF7 and trophin signalling pathway’, ‘cell–cell communication’ TamR cells (Fig. 5A). Down-regulation of phosphory- and ‘insulin signalling pathway’ (Table 3). The list of lated Yes was confirmed by blotting with a phospho- differentially phosphorylated proteins characteristic of specific antibody against the active site of SFKs. We these top seven pathways/processes was enriched also confirmed that Src protein expression was not for proteins involved in growth factor signalling affected by Yes siRNA (Fig. 5A). Knockdown of Yes (e.g. MAPK1/3, SHC1 and IRS1/2) and cell–matrix or reduced the cell-cycle progression of parental MCF7 cell–cell junctions (e.g. PTK2/FAK, CTNNA1 and cells in both the absence and presence of 4-hydroxy- CTNND1). tamoxifen (OHT). In the absence of OHT, knock- Protein–protein interactions among differentially down of Yes reduced the relative number of cells in expressed or tyrosine-phosphorylated proteins were S and G2/M phases by 36% and 25%, respectively, also interrogated using the Protein Interaction Net- while the number of cells in G1 phase was increased work Analysis (PINA) platform and the PhosphoSite- by 20%. In the presence of OHT, knockdown of Yes

Plus database [31–33]. Interestingly, the major reduced the relative number of cells in S and G2/M interaction ‘hubs’ involving proteins with expression phases by 53% and 45%, respectively, while the num- changes predominantly centred on down-regulated ber of cells in G1 phase was again increased by 20% proteins: PRKACA, CDK1, TK1, DCTN1 and ISG15 (Fig. 5B, and data not shown). However, knockdown (Fig. 3). The former three proteins play key roles in of Yes did not significantly affect the cell-cycle pro- intracellular signalling, cell-cycle control and DNA gression of TamR cells in either the absence or pres- synthesis, respectively, while DCTN1 functions in vesi- ence of OHT (Fig. 5B). In proliferation assays, Yes cle/organelle trafficking, and ISG15 is a ubiquitin-like knockdown had no effect on either cell type in either modifier involved in interferon signalling. Annexin A2 the presence or absence of OHT (Fig. 5C), probably (ANXA2), which was up-regulated in TamR cells, was reflecting the longer time frame for these assays com- also revealed as an interaction hub: annexin A2 has pared to the cell-cycle analyses, or different plating known roles in cancer cell motility and invasion [34]. densities. Transwell assays confirmed that TamR cells In contrast, when differentially tyrosine-phosphory- have a highly motile phenotype compared to parental lated proteins were considered, several proteins exhib- MCF7 cells [36] (Fig. 5D). Importantly, knockdown iting either increased or decreased tyrosine of Yes significantly reduced the migration of TamR phosphorylation exhibited high connectivity, and, cells to levels approaching that of parental MCF7 when both types were analysed together, a highly cells (Fig. 5D). interconnected network was revealed (Fig. 4). Major hubs at the core of this network that exhibited MARCKS is also required for the motile increased tyrosine phosphorylation were PTK2/FAK, phenotype of TamR cells CRKL, SHC1, STAT3 and MAPK1, while tyrosine phosphorylation was decreased for IRS1, BCAR1 and Given the markedly increased expression of ABI1. Yes kinase interacts with several proteins of this MARCKS in TamR cells (Fig. 2 and Table 1), we interaction network, including two of these interaction characterized the functional role of this protein using hubs (Fig. 4). The enhanced tyrosine phosphorylation two siRNAs targeting different exons of the of SHC1 (a scaffold that promotes Ras activation via MARCKS gene (Fig. 6A). In parental MCF7 cells, Grb2 binding), MAPK1 and PTK2 is consistent with knockdown of MARCKS attenuated cell-cycle pro- previous reports that link enhanced Ras/MAPK [12] gression in both the absence and presence of OHT. and PTK2/FAK [35] signalling to endocrine resis- In the absence of OHT, the effects of MARCKS tance. knockdown on the number of cells in S, G2/M and G1 phase were a 28% decrease, 3% decrease and a 12% increase, respectively. In the presence of OHT, Functional characterization of Yes in MCF7 and the effects of MARCKS knockdown on the number TamR cells of cells in S, G2/M and G1 phase were a 38% As increased site-selective phosphorylation of two decrease, 29% decrease and a 13% increase, respec- sites in Yes was amongst the largest changes in tyro- tively (Fig. 6B, and data not shown). However, no sine phosphorylation detected in TamR cells (Fig. 2 effects on proliferation were observed (Fig. 6C). In and Table 2), we determined the functional role of TamR cells, MARCKS knockdown did not affect Yes by siRNA-mediated knockdown. Two siRNAs cell-cycle progression under either condition. In pro- targeting different exons of the YES1 gene markedly liferation assays, siRNA#2 marginally sensitized the

5242 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

SRRM1 HIST1H1B PRPF19

BAG3

CBX3 UGDH CRIP2 ACADVL HSPB8 AP2A1 DCTN2

ATP1A1 HIST1H1E ATP1B1 KIF11

CDK1 BZW1 DDAH2 TK1 DCTN1 PREX1 CLDN3 STMN1 NAGK PRKACA SEC23A

ASNA1 BAG6 SSR1 ACTR1A NUMA1 ANP32E S100P ACOX1 SSR4 MGMT IARS2 EZR PSMD9

SLC9A3R1 SPTAN1 SCP2 TCP1

ANXA6 GDI2 ANXA2 ACAA2

MYOF ISG15 LYPLA1 FLNB EPPK1 PLS3 S100A10 S100A6 ANXA3 TXNRD1 PGD ANXA11

CIRBP ETFA S100A16 DBI ARPC1B LGALS1 BLVRA GLO1 CRABP2 ACAT2 PITRM1

LXN ETFB S100A14 RAD23B ACTR3 LGALS3 RTN3 UBE2L3 ABAT MGST1 CKAP4

G6PD PPP1CB GSTM3 SSNA1 SELENBP1 NR2C2AP HPRT1 PFKP BID DCXR UBE2C

Fig. 3. Protein–protein interactions among differentially expressed proteins in TamR cells. Interactions were identiﬁed using PINA and the PhosphoSitePlus database. Red circles represent up-regulated proteins and green circles represent down-regulated proteins. Connecting lines indicate known interactions, and pink arrows represent kinase–substrate relationships. cells to OHT, whilst siRNA#1, which gave a compa- either siRNA signiﬁcantly reduced the migration of rable knockdown, had no effect. In contrast, in TamR cells (P < 0.0001) (Fig. 6D). These data indi- Transwell assays, knockdown of MARCKS with cate that, while MARCKS expression is required for

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5243 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al.

PKP4

CTNND2 PARD3 PTK6

ERBB2IP CTNND1

TENC1 CTNNA1

YES1 CRKL TNK2 BCAR1

STAT3 WASL ABI1 TYK2 KRT18 INPPL1 PTK2

CDK1 FRK MAPK1 SHC1

MAPK3 Fig. 4. Protein–protein interactions among IRS1 differentially phosphorylated proteins in APP IRS2 TamR cells. Interactions were identiﬁed using PINA and the PhosphoSitePlus ATP1A1 database. Red circles represent proteins APLP2 with up-regulated phosphorylation and green circles represent proteins with down-regulated phosphorylation. Connecting lines indicate known interactions, and pink arrows represent PTTG1IP EPHA4 CALM1 G6PD kinase–substrate relationships.

the enhanced motility of TamR cells, it is not a higher in ER-negative compared to ER-positive can- major determinant of their anti-estrogen resistant cer cell lines, though these data did not reach statis- phenotype. tical signiﬁcance (Fig. 7A,B).

MARCKS mRNA and protein expression in a MARCKS expression in primary human breast breast cell line panel cancers and correlation with clinicopathological data To determine whether high expression of MARCKS was common in breast cancer cell lines, we measured Immunohistochemistry assays for MARCKS were MARCKS mRNA and protein levels in a panel of undertaken on tissue microarrays derived from 250 22 cell lines, consisting of seven ER-positive breast breast cancer patients with known clinical outcome cancer lines, 10 ER-negative lines, and five normal [37,38]. The results of a representative MARCKS or immortalized mammary epithelial cell lines immunohistochemistry assay are presented in Fig. 8A. (Fig. 7A,B). A significant correlation was observed Cytoplasmic staining was present in 101 samples between levels of MARCKS mRNA and protein (40.4%), with variable intensity (1+ to 3+). The per- expression (Fig. 7C). Both MARCKS mRNA and centage of MARCKS-positive cells ranged from 0% to protein levels were variable in ER-positive and -nega- 96%, with a mean of 14.2% and a median of 0%. tive cell lines, and overall MARCKS expression was MARCKS cytoplasmic expression (> 0%) correlated significantly higher in cancer cell lines than non- with grade 3 cancers (P < 0.0001), ER and progester- transformed cells. There was also a trend for one receptor (PR) negativity (P < 0.0001), HER2 MARCKS mRNA and protein expression to be amplification (P = 0.0014) and p53 positivity

5244 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

MCF7 TamR A C Proliferation 1.00 MCF7

siNT siNT siYes #1 siYes #2 siYes #1 siYes #2

Yes 0.10 pYes siNT Src siYes #1 Log Absorbance @ 595 nm 0.01 Actin 1357 siYes #2 Days siNT +OHT

1.00 TamR si #1 +OHT B Cell cycle S-phase si #2 +OHT MCF7 1.0 0.10

0.5 Log Absorbance @ 595 nm 0.01 1357 Days Relative no. cells in S phase Motility siNT siYes #1 siYes #2 D *** siNT + OHTsi #1 + OHTsi #2 + OHT *** TamR 60 P < 0.0001

1.0 40

0.5 20 No. migrated cells per field Relative no. cells in S phase

siNT siYes siYes siNT #1 #2 siYes #1 siYes #2 siNT + OHTsi #1 + OHTsi #2 + OHT MCF7 TamR

Fig. 5. siRNA-mediated knockdown of Yes inhibits motility of TamR cells. (A) Western blots of MCF7 and TamR cells after Yes or non- targeting (NT) siRNA transfection for 48 h. Down-regulation of Yes phosphorylation was conﬁrmed using an antibody against phosphorylated Src family kinases (Y416). Src protein expression was not affected by Yes siRNA. (B) Relative number of MCF7 and TamR cells in the S phase of the cell cycle after Yes or NT siRNA transfection for 48 h, and treatment with or without 4-hydroxytamoxifen (OHT) for a further 48 h. (C) Proliferation of MCF7 and TamR cells after Yes or NT siRNA transfection for 48 h, followed by 7 days with or without OHT treatment. (D) Migration of MCF7 and TamR cells through Transwell chambers. TamR cells were transfected with Yes or NT siRNA for 48 h, and the number of migrated cells was counted after a further 48 h. Values are means Æ standard deviation of duplicate or replicate results. The P value for comparison of the four groups by ANOVA is shown. Asterisks indicate statistically signiﬁcant differences (***P < 0.0001) by Tukey’s multiple comparison tests.

(P < 0.0001). MARCKS cytoplasmic expression was in luminal A than luminal B cancers (basal-like, signiﬁcantly higher in basal-like and HER2-positive 44.2%; HER2, 34.7%; luminal A, 2.6%; luminal B, cancers than in luminal cancers, and signiﬁcantly lower 12.4%) (Fig. 8B).

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5245 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al.

A MCF7 TamR C Proliferation 1.00 MCF7

siNT siMARCKS #1siMARCKS #2 siNT siMARCKS #1siMARCKS #2 0.10 MARCKS siNT x siMARCKS #1 Actin siMARCKS #2

Log Absorbance @ 595 nm 0.01 1357siNT +OHT Days si #1 +OHT 1.00 TamR B Cell cycle S-phase si #2 +OHT MCF7

1.0 0.10

0.5

Log Absorbance @ 595 nm 0.01 1357 Days Relative no. cells in S phase

siNT Motility siNT + OHTsi #1 + OHTsi #2 + OHT D *** siMARCKS #1siMARCKS #2 *** TamR 60 P < 0.0001

1.0 40

0.5 20 No. migrated cells per field Relative no. cells in S phase

siNT siNT siMARCKS siMARCKS #1 #2 siNT + OHTsi #1 + OHTsi #2 + OHT siMARCKS #1siMARCKS #2 MCF7 TamR

Fig. 6. siRNA knockdown of MARCKS inhibits motility of TamR cells. (A) Western blots of MCF7 and TamR cells after MARCKS or non- targeting (NT) siRNA transfection for 48 h. The lower band (marked ‘x’) represents non-speciﬁc antibody binding. (B) Relative number of MCF7 and TamR cells in the S phase of the cell cycle after MARCKS or NT siRNA transfection for 48 h, and treatment with or without 4-hydroxytamoxifen (OHT) for a further 48 h. (C) Proliferation of MCF7 and TamR cells after MARCKS or NT siRNA transfection for 48 h, followed by 7 days with or without OHT treatment. (D) Migration of MCF7 and TamR cells through Transwell chambers. TamR cells were transfected with MARCKS or NT siRNA for 48 h, and the number of migrated cells was counted after a further 48 h. Values are means Æ standard deviation of duplicate or replicate results. The P –value for comparison of the four groups by ANOVA is shown. Asterisks indicate statistically signiﬁcant differences (***P < 0.0001) by Tukey’s multiple comparison tests.

Correlation of MARCKS expression with patient cohort was divided into patients with positive outcome (n = 101) and negative (n = 149) cytoplasmic MARCKS expression for subsequent survival analy- As the median percentage of cells positive for ses. In our breast cancer cohort, positive MARCKS MARCKS cytoplasmic expression was 0%, the

5246 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

A Normal ER +ve ER –ve

P = 0.010 P = 0.068 20 12 6 15 8 4 10 4 2 5 Relative MARCKS mRNA level

Relative MARCKS mRNA level Normal CancerRelative MARCKS mRNA level ER +ve ER –ve T47D BT-20 MCF7 BT474 BT549 BT483 SKBR3 ZR75-1 Hs578T MCF10A HMEC 184 HMEC 219-4 MDA-MB-361 MDA-MB-134 MDA-MB-157 MDA-MB-436 MDA-MB-330 MDA-MB-453 MDA-MB-231 MDA-MB-468 HMEC 184-B5 HMEC 184-A1

B Normal ER +ve ER –ve

8 P = 0.002 P = 0.118

6 3 4

4 2 2 2 1 Relative MARCKS protein level

Relative MARCKS protein level Normal Cancer ER +ve ER –ve Relative MARCKS protein level T47D MCF7 BT-20 BT483 BT474 BT549 SKBR3 ZR75-1 Hs578T MCF10A HMEC 184 HMEC 219-4 MDA-MB-134 MDA-MB-330 MDA-MB-361 MDA-MB-157 MDA-MB-436 MDA-MB-468 MDA-MB-453 MDA-MB-231 HMEC 184-B5 HMEC 184-A1

C r = 0.511 8 P = 0.015

Normal

4 ER +ve

ER –ve

Relative MARCKS protein level 10 20 Relative MARCKS mRNA level

Fig. 7. MARCKS expression in a panel of 22 breast cell lines. (A) MARCKS mRNA expression was quantified by real-time PCR, and all levels are expressed relative to the cell line HMEC 219-4. The mean relative levels of normal cell lines, all cancer cell lines, ER-positive and ER-negative cell lines are shown. (B) MARCKS protein expression was determined by western blotting, normalized against actin levels, and expressed relative to the cell line HMEC 219-4. The mean relative levels in normal cell lines, all cancer cell lines, ER-positive and ER- negative cell lines are shown. Error bars represent SEM. Student’s t test was performed to analysis differences between means. (C) Correlation between MARCKS mRNA and protein levels in all cell lines. The relationship was assessed using Pearson’s correlation (r). expression correlated with poor prognosis for breast prognostic significance in a resolved model that cancer-specific death in univariate Kaplan–Meier included lymph node positivity, PR positivity, HER2 analysis (P < 0.0001, Fig. 8C) and Cox proportional amplification and MARCKS positivity (Table 4). hazards analysis (hazard ratio 3.49, 95% confidence In the ER-positive sub-group of patients (n = 171), interval 1.943–6.293, P < 0.0001). In multivariate positive MARCKS expression again predicted poor analysis, MARCKS positivity retained independent outcome in univariate Kaplan–Meier analysis

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5247 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al.

A MARCKS - negative MARCKS - 2+ and 3+ MARCKS - 3+

B *** *** *** *** 40

** 20

Fig. 8. MARCKS expression in primary

Mean cellular cytoplasmic % MARCKS human breast cancers and correlation with BasalHER2 Luminal A Luminal B Unclassified patient survival. (A) Representative images n = 29n = 21 n = 101 n = 76 n = 14 of MARCKS-negative and -positive breast cancer specimens. Some stromal staining Whole cohort ER-positive is observed. (B) MARCKS levels in breast C Breast Cancer Specific death D Breast Cancer Specific death cancer sub-groups. Specimens were HR = 3.94 HR = 3.91 scored as the percentage of positively 1 1 P P < 0.0001 = 0.0005 staining cells, and values are .8 .8 means Æ SEM for each sub-group. Asterisks indicate statistically significant .6 .6 differences (**P < 0.005, ***P < 0.0001) .4 .4 MARCKS = 0 (n = 149) MARCKS = 0 (n = 124) by Fisher’s least significant difference. – .2 MARCKS > 0 (n = 101) .2 MARCKS > 0 (n = 47) (C, D) Kaplan Meier curves for breast Cumulative survival Cumulative survival cancer-specific death in the patient cohort (C) and in ER-positive patients (D). Kaplan– 0 4080 120 160 0 4080 120 160 Meier P values and Cox hazard ratios (HR) Time (months) Time (months) are shown.

(P = 0.0005, Fig. 8D) and Cox proportional hazards of research. In this study, we have addressed this issue analysis (hazard ratio 3.914, 95% confidence interval by analysis of the proteome and phosphoproteome of 1.708–9.970, P = 0.0013), and also independently pre- tamoxifen-resistant MCF7 cells. These novel data have dicted death in a final resolved multivariate model that permitted identification of previously unknown altera- included HER2 amplification, PR positivity and tions in signalling networks that directly contribute to MARCKS positivity (Table 4). the aggressive nature of tamoxifen-resistant disease. Use of KM plotter [39], which integrates publically Our network-level analysis of protein expression available gene expression databases and clinical data, and signalling provides important insights into regula- revealed a significant association between high tory mechanisms associated with the development of MARCKS expression and decreased relapse-free sur- tamoxifen resistance. Some insights are novel, while vival for ER-positive breast cancer patients treated other findings support previous work but provide a with tamoxifen (P = 0.0127, Fig. S1). global context for particular regulatory events for the first time. First, analysis of protein expression changes Discussion indicated a strong and highly significant enrichment for the pathway term ‘metabolism’. This finding is of Deciphering the molecular mechanisms by which particular interest given the resurgence of interest in ER-positive breast cancer cells evade the inhibitory cancer cell metabolism and the possibility that the effects of endocrine therapy remains an essential field study of metabolism may open new avenues for

5248 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

Table 4. Cox proportional hazards analyses for breast cancer cell migration [41]. This suggests that MK01/MAPK1 specific death in the whole cohort and in ER-positive patients. HR, may contribute to the enhanced motility of TamR hazard ratio; CI, confidence interval. cells. A final finding is that, while the tyrosine phos- Variable HR 95% CI P phorylation of many proteins changes in a manner consistent with their known roles in signalling and Whole cohort endocrine resistance, the expression or phosphoryla- Grade > 2 0.915 0.368–2.276 0.8490 Size > 20 mm 1.033 0.573–1.863 0.9150 tion of others is altered in a seemingly counter-intui- Lymph node+ 3.215 1.707–6.054 0.0003 tive manner. Examples are the reduced expression of ER+ 0.653 0.325–1.315 0.2330 PRKACA, and the lower tyrosine phosphorylation of PR+ 0.322 0.143–0.725 0.0062 RET, in the tamoxifen-resistant cells, as both protein HER2+ 2.664 1.430–4.965 0.0020 kinase A (PKA)- and RET-mediated signalling are – MARCKS high 1.903 0.857 4.226 0.1141 implicated in endocrine resistance in breast cancer Resolved model cells [42,43]. A possible explanation is that our global Lymph node+ 2.937 1.590–5.426 0.0006 PR+ 0.254 0.125–0.517 0.0002 approach not only identifies ‘drivers’ of tamoxifen HER2+ 2.634 1.434–4.838 0.0018 resistance in this model, but also detects compensa- MARCKS high 1.960 1.049–3.662 0.0349 tory changes within the signalling network, which ER-positive patients occur in response to negative feedback mechanisms. Grade > 2 0.889 0.278–2.837 0.8419 For example, PI3K/Akt signalling is known to sup- > – Size 20 mm 0.803 0.328 1.965 0.6309 press expression of several receptor tyrosine kinases Lymph node+ 1.207 0.496–2.938 0.6791 in cancer cells [44]. An important implication of these PR+ 0.302 0.122–0.746 0.0094 HER2+ 5.344 1.956–14.597 0.0011 data is that particular ‘downstream’ signalling effec- MARCKS high 2.661 0.895–7.912 0.0784 tors (such as MAPK1) may provide more reliable Resolved model biomarkers of endocrine sensitivity than ‘upstream’ PR+ 0.289 0.121–0.689 0.0051 signal transducers that are susceptible to modulation HER2+ 4.824 1.931–12.053 0.0008 by negative feedback regulation. – MARCKS high 2.462 1.018 5.957 0.0456 Our data demonstrating increased phosphorylation of the Ras/MAPK signalling pathway components SHC1 and MAPK1 in TamR cells are consistent with therapeutic intervention. In this context, it has a previous report describing a causative role for recently been reported that mitochondrial activity in enhanced EGFR/ERBB2/MAPK signalling in anticancer cells drives tamoxifen resistance, and that this estrogen resistance in this model [9]. However, func- may be overcome by agents that target this activity, tional characterization of two additional candidates such as metformin [40]. Further functional interroga- identified by our study, Yes and MARCKS, revealed tion of the metabolic pathways perturbed in our that neither protein was critical for resistance to the tamoxifen-resistant model may identify additional anti-proliferative effects of tamoxifen. Instead, both such therapeutic strategies. Second, our data indicate proteins were required for the highly motile phenotype that altered signalling at cell–matrix and cell–cell of TamR cells. Interestingly, a recent proteomic study junctions is associated with acquisition of tamoxifen of independently derived tamoxifen-resistant MCF7 resistance, and that these changes may contribute to cells determined that they are characterized by a protein a more motile cellular phenotype. While consistent expression signature associated with altered cytoskele- with previously published work [28], the likely ton dynamics and an enhanced migratory capacity involvement of Yes is a new finding, and is discussed [28]. Proteins up-regulated in that study as well as our in more detail below. Third, our data support a key own include S100P, MARCKS, S100A10 and annex- role for Ras/MAPK and phosphatidyl inositol in A2, and roles in cancer cell motility or invasion 3-kinase (PI3K)/Akt signalling in the development of have previously been described for each [28,45–48]. tamoxifen resistance. These data are also consistent These data indicate that increased migratory potential with the literature [2], but our work provides net- reproducibly accompanies the acquisition of tamoxifen work-level insights into how these pathways are per- resistance, at least in MCF7 cells. As enhanced cell turbed. In addition, our finding that signalling by migration is a known contributor to the metastatic MK01/MAPK1, but not MK03/MAPK3, is enhanced process, this indicates that evolution of two pheno- in TamR cells is of interest in light of a recent report types associated with poor outcome, specifically resis- describing a selective role for the former kinase in tance to anti-proliferative effects and increased mediating hepatocyte growth factor (HGF)-induced metastatic potential, may occur concurrently. A related

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5249 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al. process occurs in non-small cell lung cancer, where poor outcome after tamoxifen treatment in ER-posi- development of resistance to EGFR tyrosine kinase tive patients. While the relationship between inhibitors is associated with epithelial-to-mesenchymal MARCKS and clinical outcome may reflect a pro-met- transition [49]. Characterization of the regulatory astatic role for MARCKS, rather than a role in pro- changes that underpin these events may identify novel moting resistance to the anti-proliferative effects of strategies for therapeutic intervention, as well as tamoxifen, our work identifies MARCKS as a poten- biomarkers that aid patient stratification for optimal tially useful prognostic biomarker, particularly for ER- therapy. positive patients. Expression of both total MARCKS and MARCKS Increased expression and activity of Yes is reported phosphorylated on S152/6 was elevated in TamR cells. in 50% of colorectal cancers [59], where it enhances MARCKS is a major protein kinase C substrate that cellular motility [60]. However, its role in breast cancer binds actin filaments and associates with the plasma is poorly understood. Previously, enhanced activation membrane, regulating cytoskeletal dynamics and of SFKs in TamR cells was demonstrated by western influencing cell adhesion, cell shape and phagocytosis blotting with a phospho-specific antibody directed [50]. Upon serine phosphorylation of its effector against the conserved activation loop of SFKs [55]. domain by protein kinase C, MARCKS is released However, it was unclear from this study which SFKs from the membrane and translocates to the cytoplasm, were responsible for this observation. While we cannot altering the stability of the cytoskeleton and enhancing rule out enhanced activation of Src itself in the TamR cell motility [48,51,52]. Little has been documented cells, our combined phosphoproteomic, western blot- about an intracellular signalling role for cytoplasmic ting and knockdown approaches clearly indicate that MARCKS, although a role in insulin-dependant Yes contributes to enhanced SFK activity in these metabolism of phosphatidylinositol 4,5-bisphosphate cells. In addition, previous studies demonstrated that in endothelial cells has been described [53], and phos- administration of the tyrosine kinase inhibitor phorylated MARCKS interacts with Tob (transducer AZD0530, which has broad activity against SFKs [61], of ErbB2), decreasing the inhibitory interaction of Tob attenuated migration and invasion of these cells, but with ErbB2 and thereby promoting ErbB2-mediated did not affect cell proliferation [55]. Our study signal transduction [54]. Given the known role of indicates that Yes is a key contributor to the pro- EGFR/ERBB2 signalling in promoting the motility of migratory role of SFKs in TamR cells, and the pro- TamR cells [55], this mechanism may contribute to the tein-protein interaction networks characterised in this pro-migratory role identified for MARCKS. As TamR paper highlight known associations between Yes and cells also exhibit enhanced invasive potential [36], fur- PTK2/FAK d-catenin (CTNND1) and ABI1, each of ther investigation is required to reveal whether which has documented roles in cancer cell motility or MARCKS contributes to the enhanced invasion as tumorigenesis [62–64]. Consequently, it will be interest- well as the motility of these cells; a role for MARCKS ing to further characterize the role of Yes and other in the invasion of glioblastoma tumour cells has been individual SFKs in regulating migration and invasion described previously [51]. Over-expressing MARCKS in this and other breast cancer models, and to deter- in MCF7 cells and other cell models may also further mine their association with patient prognosis. our understanding of the role of MARCKS in endocrine resistance. While MARCKS has been implicated Experimental procedures in the pathology of various cancers, including hepatocellular carcinoma [56], glioblastoma [51,57], cholan- Cell lines giocarcinoma [52], prostate cancer [58] and breast cancer [54], to our knowledge, this study is the first to The human breast cancer cell line MCF7 was a gift from characterize MARCKS protein expression in a large AstraZeneca (Wilmington, DE, USA). The tamoxifen-resis- breast cancer patient cohort. Here, we reveal that tant (TamR) cell line was derived from these MCF7 cells as MARCKS positivity correlates with negative prognos- previously reported [9], and MCF7 and TamR cells were tic parameters, including ER negativity, and indepen- routinely maintained as previously described [9]. The cell dently predicts patient death. Furthermore, MARCKS lines BT-20, BT549, MDA-MB-468, MDA-MB-231, positivity independently predicts poor outcome in the Hs578T, MDA-MB-453, MDA-MB-330, SKBR3, MDA- ER-positive sub-group of patients, and these data are MB-436, MDA-MB-157, BT474, MDA-MB-134, ZR75-1, corroborated by use of the KM plotter [39], which BT483, MDA-MB-361, T47D, HMEC 184-A1, shows, using publically available information, that HMEC 219-4, MCF10A, HMEC 184 and HMEC 184-B5 were obtained and maintained as previously described [65]. high MARCKS gene expression also correlates with

5250 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

Stable isotopic labelling using amino acids in electrospray analysis in data-dependent acquisition mode À culture (SILAC) (250 nL min 1 over 30 min). Survey scans in the mass range of 350–1750 m/z were acquired using the Orbitrap Cells were SILAC-labelled by culturing in phenol-red free mass spectrometer, with a resolution of 60 000 at m/z 400 RPMI-1640 (Life Technologies, Mulgrave, VIC, Australia) and with lock mass enabled. The top 15 most intense ions in which the natural lysine and arginine were replaced by (ion selection threshold > 5000 counts) with charge states 13 15 the heavy isotope-labelled amino acids C6 N4-L-arginine ≥ 2 were sequentially isolated and further subjected to MS/ 13 15 (Arg10) and C6 N2-L-lysine (Lys8) (Silantes GmBH, MS fragmentation within the linear ion trap using colli- Munich, Germany), and which was supplemented with 5% sion-induced dissociation (30% relative collision energy). dialysed fetal bovine serum, 4 mM glutamine and MS/MS spectra were accumulated with an activation time l Á À1 10 g mL insulin. Cells were cultured for approximately of 30 ms at a target value of 30 000 ions, and ions were six doublings in the SILAC medium to reach complete excluded from further selection for 15 s. labelling of > 97%. Two biological replicates were created by reversing the cell types that were labelled with the heavy and light isotopes. Cells were lysed in urea lysis buffer In-gel digestion (2.5 mM sodium pyrophosphate, 1 mM B-glycerol phos- In-gel digestions were performed as previously described phate, 1 mM sodium orthovanadate, 1 mM tris (2-carboxy- [62]. ethyl) phosphine (TCEP), 1 mM EDTA, 8 M urea and 20 mM HEPES), sonicated, and then iodoacetamide was added to 100 mM. Lysates were centrifuged at 18 000 g Protein identification at 4 °C. Protein concentration was determined by the Bradford assay (Thermo Fisher Scientific, Scoresby, VIC, Mass spectrometric raw files were processed using MAX- Australia). QUANT software version 1.1.1.25 (http://www.maxquan- t.org), which employs the Andromeda algorithm for peak picking and processing as well as protein identification and Protein expression and phosphoproteomic quantification [66]. Extracted peak lists were searched profiling by mass spectrometry against a Homo sapiens Swiss-Prot sequence database (2010_10) containing 35 052 entries and a proportionately For protein expression profiling, equal amounts of lysate sized reversed database for generation of false discovery from TamR and MCF7 cells labelled with heavy and light rates. The following search parameters were selected: fixed isotopes, respectively, were mixed, and proteins were sepa- cysteine carbamidomethylation modification, variable methi- rated by SDS/PAGE, followed by in-gel digestion with onine oxidation modification, variable protein N-acetyla- trypsin, extraction of peptides and MS analysis. For phos- tion, variable phosphorylation of serine, threonine and phoproteomic profiling, equal amounts of lysate from iso- tyrosine, and a minimum peptide length of six amino acids. tope-labelled MCF7 and TamR cells were mixed, and Up to two missed cleavages were allowed. The multiplicity 40 mg were subjected to sequential phosphotyrosine pep- of experiments was set to 2, and the SILAC labels Arg10 tide immunoprecipitation with pY100 antibody (Cell Sig- and Lys8 were selected as variable modifications, with the naling Technology, Danvers, MA, USA) and pY20 minimum peptide count for protein quantification set to 1. antibody (BD Biosciences, Sparks, MD, USA), essentially The initial first search mass tolerance was 20 p.p.m. for pre- as previously described [29]. Immunoprecipitated peptides cursor ions and 0.5 Da for fragment ions, with individual- were dissolved in a buffer containing formic acid (1%), ace- ized peptide mass tolerances used for the subsequent tonitrile (2%) and heptafluorobutyric acid (0.05%). For searches. The false discovery rate was limited to 1% for both MS analyses, dissolved peptides were separated by nano- protein and peptide identifications, and ‘match between LC using an Ultimate 3000 HPLC and autosampler system runs’ was enabled with default settings. (Thermo Fisher Scientific), and mass spectra were acquired on either a LTQ-FT-ICR (Thermo Fisher Scientific) (using conditions described previously [29]) or an Orbitrap Velos Protein identification and differential expression mass spectrometer (Thermo Fisher Scientific). Briefly, sam- selection ples were desalted using a micro C18 pre-column (500 lm 9 2 mm column, Michrom Bioresources) with 2% Peptide and protein identifications were filtered based on acetonitrile in 0.05% trifluoroacetic acid at 15 lLÁminÀ1 the following criteria: for all protein identifications, the < for 4 min, followed by loading of the sample onto a Magic false discovery rate was limited to 1% and a maximum posterior error probability of 0.1 (10%) was applied. For C18 column (75 lm 9 10 cm column; particle size, 5 lm; pore size, 200 A; Bruker-Michrom, Auburn, CA, USA). pY site identifications, the false discovery rate was limited < Using increasing concentrations of acetonitrile, samples to 1% and a site localization probability minimum of were eluted, and the nano-eluent was subject to nano-flow 0.75 (75%) was set. Proteins that exhibited a change in

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5251 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al. expression ≥ 1.5-fold from both replicate SILAC experi- Cell-cycle analysis ments were considered as differentially expressed. pY sites 9 5 exhibiting ≥ 1.2-fold changes in levels in both SILAC Forty-eight hours after transfection, 8 10 cells were experiments were also considered as differentially regulated. seeded in 10 cm plates and incubated in the presence or absence of OHT (100 nM) (Sigma-Aldrich) for a further 48 h. Cells were then ﬁxed in 70% ethanol, stained using Pathway enrichment and protein–protein propidium iodide, and cell-cycle analysis was performed by interaction network analyses ﬂow cytometry.

KOBAS [67,68] was used to perform pathway enrichment analysis. The hypergeometric test was selected to test statis- Proliferation assays tical enrichment of KEGG (Kyoto Encyclopedia of Genes MCF7 and TamR cells were seeded into 24-well plates 48 h and Genomes, http://www.genome.jp/kegg/) and Reactome post-transfection at a density of 2 9 104 cells per well, and (http://www.reactome.org) pathways, and the P values were cultured in the presence or absence of OHT (100 nM). The corrected for multiple comparisons [69]. The protein–pro- relative cell number was assessed at days 1, 5 and 7 using a tein interactions among proteins of interest were obtained crystal violet-based colorimetric assay. Briefly, cells were from the Protein Interaction Network Analysis (PINA) fixed with trichloroacetic acid (16% v/v) for 30 min, and platform [31,33], and substrate–kinase relationships were then rinsed thoroughly with dH O. Cells were stained using downloaded from the PhosphoSitePlus database [32]. A 2 Diff-Quik Stain 2 (1 : 10 dilution; Bio-Scientifica, Kirraq- Cytoscape [70] plug-in for PINA (H.C. Lee, M. Pinese, wee, NSW, Australia), rinsed and dried. The remaining M.J. Cowley, The Kinghorn Cancer Centre (TKCC), Gar- stain was then dissolved in 300 lL acetic acid (10% v/v) van Institute of Medical Research (GIMR), NSW, Austra- and quantified at 595 nm. Values are the mean absorbance lia; S. M. Grimmond, Queensland Centre for Medical per sample, corrected for background. Genomics, Institute for Molecular Bioscience, The Univer- sity of Queensland, QLD, Australia; A.V. Biankin, R.J.D, J.W, TKCC, GIMR, NSW, Australia, unpublished results) Migration assays was used for visualization of networks. Porous 24-well membrane insert (6.5 mm diameter/8 lm pore membrane) (BD Biosciences) were basally coated with Western blotting fibronectin (1% v/v in RPMI-1640 medium) (Sigma- Aldrich) overnight at 4 °C. Inserts were then washed with Cell lysates were prepared and immunoblotted as previ- NaCl/P and allowed to air dry before being transferred to ously described [71]. Antibodies against the following pro- i a well containing 650 lL of medium. MCF7 and TamR teins were used: BID (BD Biosciences), Src, IRS1 cells were then seeded (in triplicate) into the insert’s (Millipore, Kilsyth, Vic, Australia), MARCKS, STAT3 apical chamber 48 h post-transfection (200 lL, 2.5 9 105 (Thermo Fisher Scientific), Yes (WAKO Biochemicals, cellsÁmLÀ1). Following a further 48 h culture, apical med- Richmond, VA, USA), Ezrin, MAPK1/3, phospho- ium was then aspirated, and non-migratory cells attached MAPK1/3 T202/Y204, phosphoSrc family Y416, phospho- to the apical surface of the membrane were removed with a STAT3 Y705, phosphoMARCKS S152/156 (Cell Signaling cotton bud. Cells attached to the basal membrane were Technology), and phosphoIRS1 Y612 (Life Technologies). fixed with formaldehyde (3.7% v/v) for 10 min. The basal Blots were checked for equal loading by re-probing with an membrane was then rinsed with NaCl/P and submerged antibody against actin (Sigma-Aldrich, Sydney, NSW, Aus- i into crystal violet (0.5% w/v) for 30 min. Inserts were tralia). rinsed thoroughly, dried and visualized at 109 magnifica- tion. The number of migratory cells present in five fields of siRNA-mediated knockdown view was recorded per insert, and data are presented as the mean count/field. Small interfering RNAs (siRNAs) specific for MARCKS (#1, ON-TARGETplus MARCKS #8-J004772-08; #2, ON-TARGETplus MARCKS #9-J004772-09), Yes1 (#1, Real-time quantitative PCR ON-TARGETplus Yes1 #12-J-003184-12; #2,ON-TAR- Total RNA was isolated using an RNeasy kit (Qiagen, Chad- GETplus Yes1 #13-J-003184-13) and non-targeting controls stone Centre, Vic, Australia) and was reverse-transcribed (ON-TARGETplus Non Targeting siRNA #1-D-001810-01) using the Reverse Transcription System (Promega, Alexan- were purchased from Thermo Fisher Scientific. Cells were dria, NSW, Australia) according to the manufacturer’s transfected using Lipofectamine 2000 (Life Technologies) in instructions. Real-time PCR was performed using an ABI the presence of the siRNAs according to the manufac- Prism 7900HT sequence detection system (Life Technologies) turer’s instructions.

5252 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance using inventoried Taq-Man probes for MARCKS (Life (Dako Australia) for 10 min. Slides were then rinsed in Technologies). RPLPO (large ribosomal protein) (Life Tech- water and counter-stained with haematoxylin, dehydrated nologies) was used as an internal loading control. Data anal- through graded ethanol, cleared in xylene and mounted. yses were performed using the DCt method, as previously All staining was assessed by an experienced breast described [65]. Fold changes in gene expression were calcu- pathologist, blinded to all clinical, molecular and outcome lated relative to normal breast cell lines. data. Staining for MARCKS was assessed in the epithelium (cytoplasmic), and was scored for the percentage of positively staining cells and the intensity of staining using a Patient characteristics semi-quantitative method (0, negative; 1+, weak; 2+, mod- Formalin-fixed, paraffin-embedded tissue blocks were erate; 3+, strong). All data were entered into CanSto, the obtained from 292 patients diagnosed with invasive ductal data management program of the Kinghorn Cancer Centre breast carcinoma between 1992 and 2002 at St Vincent’s (TKCC) and the Garvan Institute of Medical Research, Public and Private Hospitals, Sydney, Australia. Tissue Sydney, Australia. microarrays were constructed with two to six 1 mm cores per patient tumour. Prior approval for this study was obtained Statistical analyses from the Human Research Ethics Committee of St Vincent’s Hospital. Clinical details have been published [37,38]. Briefly, The association of MARCKS protein expression with clini- the median follow-up time was 64 months (range 0– copathological variables was tested using a v2 test in con- 152 months). Forty per cent of tumours were > 20 mm, 45% tingency tables, and the relationship between its expression were grade > 2, 43% were lymph node-positive, 68% were as a continuous variable and intrinsic subtype was explored ER-positive, 57% were PR-positive and 18% were HER2- using analysis of variance (ANOVA) with Fisher’s post hoc positive by fluorescence in situ hybridization (ratio of HER2: test of significance. Kaplan–Meier and the Cox propor- CEP17 (chromosome 17 centromere) > 2.2). The median age tional hazards model were used for univariate analysis and was 54 years, and patients were treated with endocrine multivariate analyses. Those factors that were prognostic in therapy (49%), chemotherapy (38%) or both (24%). The univariate analysis were then assessed in a multivariable intrinsic molecular subtype was assessed by immunohisto- model, with backwards stepwise elimination of redundant chemistry for ER, PR, CK5/6, EGFR and HER2, and fluo- variables, to identify those that were independently prog- rescence in situ hybridization for HER2 was performed as nostic and those that were the result of confounding vari- previously described [37]. p53 staining was also performed as ables. All analyses were performed using STATVIEW 5.0 previously described [65]. (http://www.statview.com) or GRAPHPAD PRISM 6.0 (http:// www.graphpad.com). MARCKS immunohistochemistry and scoring Acknowledgements Sections 4 lm thick were cut from each tissue microarray, mounted on SuperFrost Plus glass slides (Thermo Fisher This work was supported by the Australian National Scientific) and baked for 2 h at 63 °C, then dewaxed by Health and Medical Research Council (program grant passage through xylene (two 5 min washes), cleared and 535903 to R.J.D., E.A.M. and R.L.S., fellowship rehydrated in graded alcohol (100, 95 and 70%, 2 min 427601 to R.L.S., and grant 535914 to R.J.D.), the each), ending in a distilled water wash (1 min). Antigen Cancer Institute of New South Wales (Translational retrieval was performed using Target Retrieval Solution Program Grant 10/TPG/1-04 to R.L.S., E.A.M., (pH 6.0) (Dako Australia, Campbellfield, Vic, Australia) in E.K.A.M., S.A.O.), the Sydney Catalyst Translational ° a Pascal pressure chamber (Dako Australia) at 125 C for Research Centre (grant 11/TRC/1-02 to R.L.S. and ° 60 s, then 95 C for 10 s, followed by cooling gently for fellowship 10/CRF/1-07 to S.A.O.), the Australia and 15 min in a running water bath and washing in distilled New Zealand Breast Cancer Trials Group (R.L.S. and water for 1 min. Using the Dako Autostainer Plus Link S.A.O.), the Australian Cancer Research Foundation (Dako Australia), endogenous peroxidase activity was elim- (R.L.S., E.A.M., R.J.D. and A.B.), the Sydney Breast inated using 3% hydrogen peroxide for 5 min, and the tis- Cancer Foundation (S.A.O.), the RT Hall Trust sue microarrays were blocked in Protein Block, Serum-Free (Dako Australia) for 10 min. Slides were incubated with (R.L.S) and the Petre Foundation (R.L.S.). R.L. is a rabbit monoclonal antibody against MARCKS Lundbeck Foundation Fellow, and is also supported (0.12 lgÁmLÀ1; Cell Signaling Technology) for 60 min at by a Sapere Aude Starting Grant from the Danish room temperature. Following rinsing in Wash Buffer Council for Independent Research and a Career Devel- (Dako Australia), slides were incubated in Envision+ Per- opment Award from the Human Frontier Science oxidase Labelled Polymer, Anti-Rabbit (Dako Australia) Program. F.H. is in receipt of funding from the for 30 min, rinsed and incubated in DAB+ chromagen Bildungsministerium fur€ Bildung und Forschung

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5253 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al.

(03Z1CN21). The authors wish to thank Paul Timp- of estrogen receptor a: a new model for anti-estrogen son, Laurent Pangon and Timothy Molloy for their resistance. J Biol Chem 276, 9817–9824. critical analysis of this manuscript. 12 Gee JM, Robertson JF, Ellis IO & Nicholson RI (2001) Phosphorylation of ERK1/2 mitogen-activated protein kinase is associated with poor response to anti- References hormonal therapy and decreased patient survival in – 1 Early Breast Cancer Trialists’ Collaborative Group clinical breast cancer. Int J Cancer 95, 247 254. (2005) Effects of chemotherapy and hormonal therapy 13 Shoman N, Klassen S, McFadden A, Bickis MG, for early breast cancer on recurrence and 15-year Torlakovic E & Chibbar R (2005) Reduced PTEN survival: an overview of the randomised trials. Lancet expression predicts relapse in patients with breast 365, 1687–1717. carcinoma treated by tamoxifen. Mod Pathol 18, – 2 Musgrove EA & Sutherland RL (2009) Biological 250 259. determinants of endocrine resistance in breast cancer. 14 Morgan L, Gee J, Pumford S, Farrow L, Finlay P, Nat Rev Cancer 9, 631–643. Robertson J, Ellis I, Kawakatsu H, Nicholson R & 3 Johnston SR (2010) New strategies in estrogen Hiscox S (2009) Elevated Src kinase activity attenuates receptor-positive breast cancer. Clin Cancer Res 16, tamoxifen response in vitro and is associated with poor – 1979–1987. prognosis clinically. Cancer Biol Ther 8, 1550 1558. 4 Johnston SR (2006) Clinical efforts to combine endocrine 15 Chanrion M, Negre V, Fontaine H, Salvetat N, Bibeau agents with targeted therapies against epidermal growth F, MacGrogan G, Mauriac L, Katsaros D, Molina F, factor receptor/human epidermal growth factor Theillet C, et al. (2008) A gene expression signature receptor 2 and mammalian target of rapamycin in breast that can predict the recurrence of tamoxifen-treated – cancer. Clin Cancer Res 12, 1061s–1068s. primary breast cancer. Clin Cancer Res 14, 1744 1752. 5 Dodwell D, Wardley A & Johnston S (2006) 16 Loi S, Haibe-Kains B, Desmedt C, Lallemand F, Tutt Postmenopausal advanced breast cancer: options for AM, Gillet C, Ellis P, Harris A, Bergh J, Foekens JA, therapy after tamoxifen and aromatase inhibitors. et al. (2007) Definition of clinically distinct molecular Breast 15, 584–594. subtypes in estrogen receptor-positive breast carcinomas – 6 Bertelli G & Paridaens R (2006) Optimal sequence of through genomic grade. J Clin Oncol 25, 1239 1246. hormonotherapy in advanced breast cancer. Curr Opin 17 Ma XJ, Wang Z, Ryan PD, Isakoff SJ, Barmettler A, Oncol 18, 572–577. Fuller A, Muir B, Mohapatra G, Salunga R, Tuggle 7 Arpino G, Green SJ, Allred DC, Lew D, Martino S, JT, et al. (2004) A two-gene expression ratio predicts Osborne CK & Elledge RM (2004) HER-2 clinical outcome in breast cancer patients treated with – amplification, HER-1 expression, and tamoxifen tamoxifen. Cancer Cell 5, 607 616. response in estrogen receptor-positive metastatic breast 18 Miller TW, Balko JM, Ghazoui Z, Dunbier A, cancer: a Southwest Oncology Group study. Clin Anderson H, Dowsett M, Gonzalez-Angulo AM, Mills Cancer Res 10, 5670–5676. GB, Miller WR, Wu H, et al. (2011) A gene expression 8 Massarweh S, Osborne CK, Creighton CJ, Qin L, signature from human breast cancer cells with acquired Tsimelzon A, Huang S, Weiss H, Rimawi M & Schiff R hormone independence identifies MYC as a mediator of – (2008) Tamoxifen resistance in breast tumors is driven antiestrogen resistance. Clin Cancer Res 17, 2024 2034. by growth factor receptor signaling with repression of 19 Musgrove EA, Sergio CM, Loi S, Inman CK, classic estrogen receptor genomic function. Cancer Res Anderson LR, Alles MC, Pinese M, Caldon CE, 68, 826–833. Schutte J, Gardiner-Garden M, et al. (2008) 9 Knowlden JM, Hutcheson IR, Jones HE, Madden T, Identification of functional networks of estrogen- and Gee JM, Harper ME, Barrow D, Wakeling AE & c-Myc-responsive genes and their relationship to Nicholson RI (2003) Elevated levels of epidermal response to tamoxifen therapy in breast cancer. PLoS growth factor receptor/c-erbB2 heterodimers mediate an One 3, e2987. autocrine growth regulatory pathway in tamoxifen- 20 Paik S, Shak S, Tang G, Kim C, Baker J, Cronin M, resistant MCF-7 cells. Endocrinology 144, 1032–1044. Baehner FL, Walker MG, Watson D, Park T, et al. 10 Assender JW, Gee JM, Lewis I, Ellis IO, Robertson JF (2004) A multigene assay to predict recurrence of & Nicholson RI (2007) Protein kinase C isoform tamoxifen-treated, node-negative breast cancer. N Engl – expression as a predictor of disease outcome on J Med 351, 2817 2826. endocrine therapy in breast cancer. J Clin Pathol 60, 21 Zhang Y, Sieuwerts AM, McGreevy M, Casey G, 1216–1221. Cufer T, Paradiso A, Harbeck N, Span PN, Hicks DG, 11 Campbell RA, Bhat-Nakshatri P, Patel NM, Crowe J, et al. (2009) The 76-gene signature defines Constantinidou D, Ali S & Nakshatri H (2001) high-risk patients that benefit from adjuvant tamoxifen – Phosphatidylinositol 3-kinase/AKT-mediated activation therapy. Breast Cancer Res Treat 116, 303 309.

5254 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

22 Jansen MP, Foekens JA, van Staveren IL, Dirkzwager- 31 Cowley MJ, Pinese M, Kassahn KS, Waddell N, Kiel MM, Ritstier K, Look MP, Meijer-van Gelder Pearson JV, Grimmond SM, Biankin AV, Hautaniemi ME, Sieuwerts AM, Portengen H, Dorssers LC, et al. S & Wu J (2012) PINA v2.0: mining interactome (2005) Molecular classification of tamoxifen-resistant modules. Nucleic Acids Res 40, D862–D865. breast carcinomas by gene expression profiling. J Clin 32 Hornbeck PV, Kornhauser JM, Tkachev S, Zhang B, Oncol 23, 732–740. Skrzypek E, Murray B, Latham V & Sullivan M (2012) 23 Lyng MB, Laenkholm AV, Tan Q, Vach W, PhosphoSitePlus: a comprehensive resource for Gravgaard KH, Knoop A & Ditzel HJ (2013) Gene investigating the structure and function of expression signatures that predict outcome of experimentally determined post-translational tamoxifen-treated estrogen receptor-positive, high-risk, modifications in man and mouse. Nucleic Acids Res 40, primary breast cancer patients: a DBCG study. PLoS D261–D270. One 8, e54078. 33 Wu J, Vallenius T, Ovaska K, Westermarck J, Makela 24 Huber-Keener KJ, Liu X, Wang Z, Wang Y, Freeman TP & Hautaniemi S (2009) Integrated network analysis W, Wu S, Planas-Silva MD, Ren X, Cheng Y, Zhang platform for protein–protein interactions. Nat Methods Y, et al. (2012) Differential gene expression in 6,75–77. tamoxifen-resistant breast cancer cells revealed by a 34 Wu B, Zhang F, Yu M, Zhao P, Ji W, Zhang H, Han J new analytical model of RNA-Seq data. PLoS One 7, & Niu R (2012) Up-regulation of Anxa2 gene promotes e41333. proliferation and invasion of breast cancer MCF-7 25 Nielsen TO, Parker JS, Leung S, Voduc D, Ebbert M, cells. Cell Prolif 45, 189–198. Vickery T, Davies SR, Snider J, Stijleman IJ, Reed J, 35 Hiscox S, Jordan NJ, Morgan L, Green TP & et al. (2010) A comparison of PAM50 intrinsic Nicholson RI (2007) Src kinase promotes adhesion- subtyping with immunohistochemistry and clinical independent activation of FAK and enhances cellular prognostic factors in tamoxifen-treated estrogen migration in tamoxifen-resistant breast cancer cells. receptor-positive breast cancer. Clin Cancer Res 16, Clin Exp Metastasis 24, 157–167. 5222–5232. 36 Hiscox S, Morgan L, Barrow D, Dutkowskil C, 26 Besada V, Diaz M, Becker M, Ramos Y, Wakeling A & Nicholson RI (2004) Tamoxifen Castellanos-Serra L & Fichtner I (2006) Proteomics resistance in breast cancer cells is accompanied by an of xenografted human breast cancer indicates novel enhanced motile and invasive phenotype: inhibition by targets related to tamoxifen resistance. Proteomics 6, gefitinib (‘Iressa’, ZD1839). Clin Exp Metastasis 21, 1038–1048. 201–212. 27 Umar A, Kang H, Timmermans AM, Look MP, 37 Millar EK, Anderson LR, McNeil CM, O’Toole SA, Meijer-van Gelder ME, den Bakker MA, Jaitly N, Pinese M, Crea P, Morey AL, Biankin AV, Henshall Martens JW, Luider TM, Foekens JA, et al. (2009) SM, Musgrove EA, et al. (2009) BAG-1 predicts Identification of a putative protein profile associated patient outcome and tamoxifen responsiveness in with tamoxifen therapy resistance in breast cancer. Mol ER-positive invasive ductal carcinoma of the breast. Cell Proteomics 8, 1278–1294. Br J Cancer 100, 123–133. 28 Zhou C, Zhong Q, Rhodes LV, Townley I, Bratton 38 Millar EK, Dean JL, McNeil CM, O’Toole SA, MR, Zhang Q, Martin EC, Elliott S, Collins-Burow Henshall SM, Tran T, Lin J, Quong A, Comstock CE, BM, Burow ME, et al. (2012) Proteomic analysis of Witkiewicz A, et al. (2009) Cyclin D1b protein acquired tamoxifen resistance in MCF-7 cells reveals expression in breast cancer is independent of cyclin D1a expression signatures associated with enhanced and associated with poor disease outcome. Oncogene migration. Breast Cancer Res 14, R45. 28, 1812–1820. 29 Hochgrafe F, Zhang L, O’Toole SA, Browne BC, 39 Gyorffy B, Lanczky A, Eklund AC, Denkert C, Pinese M, Porta Cubas A, Lehrbach GM, Croucher Budczies J, Li Q & Szallasi Z (2010) An online DR, Rickwood D, Boulghourjian A, et al. (2010) survival analysis tool to rapidly assess the effect of Tyrosine phosphorylation profiling reveals the signaling 22,277 genes on breast cancer prognosis using network characteristics of basal breast cancer cells. microarray data of 1,809 patients. Breast Cancer Res Cancer Res 70, 9391–9401. Treat 123, 725–731. 30 Chappell DS, Patel NA, Jiang K, Li P, Watson JE, 40 Martinez-Outschoorn UE, Goldberg A, Lin Z, Ko YH, Byers DM & Cooper DR (2009) Functional Flomenberg N, Wang C, Pavlides S, Pestell RG, involvement of protein kinase C-bII and its substrate, Howell A, Sotgia F, et al. (2011) Anti-estrogen myristoylated alanine-rich C-kinase substrate resistance in breast cancer is induced by the tumor (MARCKS), in insulin-stimulated glucose transport microenvironment and can be overcome by inhibiting in L6 rat skeletal muscle cells. Diabetologia 52, mitochondrial function in epithelial cancer cells. Cancer 901–911. Biol Ther 12, 924–938.

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5255 Phosphoproteomic analysis of tamoxifen resistance B. C. Browne et al.

41 Radtke S, Milanovic M, Rosse C, de Rycker M, cholangiocarcinoma cell migration and metastasis via Lachmann S, Hibbert A, Kermorgant S & Parker PJ the protein kinase C-dependent pathway. Cancer Sci (2013) ERK2 but not ERK1 mediates HGF-induced 101, 658–665. motility in non small cell lung carcinoma cell lines. 53 Kalwa H & Michel T (2011) The MARCKS protein J Cell Sci 126, 2381–2391. plays a critical role in phosphatidylinositol 4,5- 42 Kok M, Zwart W, Holm C, Fles R, Hauptmann M, bisphosphate metabolism and directed cell movement in Van’t Veer LJ, Wessels LF, Neefjes J, Stal O, Linn SC, vascular endothelial cells. J Biol Chem 286, 2320–2330. et al. (2011) PKA-induced phosphorylation of ERa at 54 Jin Cho S, La M, Ahn JK, Meadows GG & Joe CO serine 305 and high PAK1 levels is associated with (2001) Tob-mediated cross-talk between MARCKS sensitivity to tamoxifen in ER-positive breast cancer. phosphorylation and ErbB-2 activation. Biochem Breast Cancer Res Treat 125,1–12. Biophys Res Commun 283, 273–277. 43 Plaza-Menacho I, Morandi A, Robertson D, Pancholi 55 Hiscox S, Morgan L, Green TP, Barrow D, Gee J & S, Drury S, Dowsett M, Martin LA & Isacke CM Nicholson RI (2006) Elevated Src activity promotes (2010) Targeting the receptor tyrosine kinase RET cellular invasion and motility in tamoxifen resistant sensitizes breast cancer cells to tamoxifen treatment and breast cancer cells. Breast Cancer Res Treat 97, reveals a role for RET in endocrine resistance. 263–274. Oncogene 29, 4648–4657. 56 Masaki T, Tokuda M, Yoshida S, Nakai S, Morishita 44 Chandarlapaty S, Sawai A, Scaltriti M, Rodrik- A, Uchida N, Funaki T, Kita Y, Funakoshi F, Outmezguine V, Grbovic-Huezo O, Serra V, Majumder Nonomura T, et al. (2005) Comparison study of the PK, Baselga J & Rosen N (2011) AKT inhibition expressions of myristoylated alanine-rich kinase relieves feedback suppression of receptor tyrosine substrate in hepatocellular carcinoma, liver cirrhosis, kinase expression and activity. Cancer Cell 19,58–71. chronic hepatitis, and normal liver. Int J Oncol 26, 45 Nipp M, Elsner M, Balluff B, Meding S, Sarioglu H, 661–671. Ueffing M, Rauser S, Unger K, Hofler H, Walch A, 57 Jarboe JS, Anderson JC, Duarte CW, Mehta T, et al. (2012) S100-A10, thioredoxin, and S100-A6 as Nowsheen S, Hicks PH, Whitley AC, Rohrbach TD, biomarkers of papillary thyroid carcinoma with lymph McCubrey RO, Chiu S, et al. (2012) MARCKS node metastasis identified by MALDI imaging. J Mol regulates growth and radiation sensitivity and is a novel Med 90, 163–174. prognostic factor for glioma. Clin Cancer Res 18, 46 Sayeed S, Asano E, Ito S, Ohno K, Hamaguchi M & 3030–3041. Senga T (2013) S100A10 is required for the 58 Li T, Li D, Sha J, Sun P & Huang Y (2009) organization of actin stress fibers and promotion of cell MicroRNA-21 directly targets MARCKS and spreading. Mol Cell Biochem 374, 105–111. promotes apoptosis resistance and invasion in 47 Zhang J, Guo B, Zhang Y, Cao J & Chen T (2010) prostate cancer cells. Biochem Biophys Res Commun Silencing of the annexin II gene down-regulates the 383, 280–285. levels of S100A10, c-Myc, and plasmin and inhibits 59 Park J, Meisler AI & Cartwright CA (1993) c-Yes breast cancer cell proliferation and invasion. Saudi Med tyrosine kinase activity in human colon carcinoma. J 31, 374–381. Oncogene 8, 2627–2635. 48 Chen X & Rotenberg SA (2010) PhosphoMARCKS 60 Barraclough J, Hodgkinson C, Hogg A, Dive C & drives motility of mouse melanoma cells. Cell Signal 22, Welman A (2007) Increases in c-Yes expression level 1097–1103. and activity promote motility but not proliferation of 49 Neel DS & Bivona TG (2013) Secrets of drug resistance human colorectal carcinoma cells. Neoplasia 9, in NSCLC exposed by new molecular definition of 745–754. EMT. Clin Cancer Res 19,3–5. 61 Green TP, Fennell M, Whittaker R, Curwen J, Jacobs 50 Ramsden JJ (2000) MARCKS: a case of molecular V, Allen J, Logie A, Hargreaves J, Hickinson DM, exaptation? Int J Biochem Cell Biol 32, 475–479. Wilkinson RW, et al. (2009) Preclinical anticancer 51 Micallef J, Taccone M, Mukherjee J, Croul S, Busby J, activity of the potent, oral Src inhibitor AZD0530. Mol Moran MF & Guha A (2009) Epidermal growth factor Oncol 3, 248–261. receptor variant III-induced glioma invasion is 62 Carr HS, Zuo Y, Oh W & Frost JA (2013) Regulation mediated through myristoylated alanine-rich protein of FAK activation, breast cancer cell motility and kinase C substrate overexpression. Cancer Res 69, amoeboid invasion by the RhoA GEF Net1. Mol Cell 7548–7556. Biol 33, 2773–2786. 52 Techasen A, Loilome W, Namwat N, Takahashi E, 63 Steinestel K, Bruderlein S, Steinestel J, Markl B, Sugihara E, Puapairoj A, Miwa M, Saya H & Schwerer MJ, Arndt A, Kraft K, Propper C & Moller Yongvanit P (2010) Myristoylated alanine-rich C kinase P (2012) Expression of Abelson interactor 1 (Abi1) substrate phosphorylation promotes correlates with inflammation, KRAS mutation and

5256 FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS B. C. Browne et al. Phosphoproteomic analysis of tamoxifen resistance

adenomatous change during colonic carcinogenesis. 69 Benjamini Y & Yekutieli D (2001) The control of the PLoS One 7, e40671. false discovery rate in multiple testing under 64 Zhang JY, Wang Y, Zhang D, Yang ZQ, Dong XJ, dependency. Ann Stat 29, 1165–1188. Jiang GY, Zhang PX, Dai SD, Dong QZ, Han Y, et al. 70 Smoot ME, Ono K, Ruscheinski J, Wang PL & Ideker (2010) d-Catenin promotes malignant phenotype of T (2011) Cytoscape 2.8: new features for data non-small cell lung cancer by non-competitive binding integration and network visualization. Bioinformatics to E-cadherin with p120ctn in cytoplasm. J Pathol 222, 27, 431–432. 76–88. 71 Janes PW, Daly RJ, deFazio A & Sutherland RL 65 Roberts CG, Millar EK, O’Toole SA, McNeil CM, (1994) Activation of the Ras signalling pathway in Lehrbach GM, Pinese M, Tobelmann P, McCloy RA, human breast cancer cells overexpressing erbB-2. Musgrove EA, Sutherland RL, et al. (2011) Oncogene 9, 3601–3608. Identification of PUMA as an estrogen target gene that mediates the apoptotic response to tamoxifen in human breast cancer cells and predicts patient outcome and Supporting information tamoxifen responsiveness in breast cancer. Oncogene 30, Additional supporting information may be found in 3186–3197. the online version of this article at the publisher’s web 66 Cox J, Neuhauser N, Michalski A, Scheltema RA, site: Olsen JV & Mann M (2011) Andromeda: a peptide Fig. S1. Kaplan–Meier curve showing relapse-free sur- search engine integrated into the MaxQuant vival in tamoxifen-treated ER-positive breast cancer – environment. J Proteome Res 10, 1794 1805. patients. 67 Wu J, Mao X, Cai T, Luo J & Wei L (2006) KOBAS Table S1. List of 186 differentially expressed proteins server: a web-based platform for automated annotation in TamR versus MCF7 cells, and all proteins identified and pathway identification. Nucleic Acids Res 34, by total proteomic analysis of MCF7 and TamR cells. W720–W724. Table S2. List of 51 differentially regulated pY sites in 68 Xie C, Mao X, Huang J, Ding Y, Wu J, Dong S, Kong TamR cells, and list of all 116 pY sites identified by L, Gao G, Li CY & Wei L (2011) KOBAS 2.0: a web server for annotation and identification of enriched phosphoproteomic analysis of MCF7 and TamR cells. pathways and diseases. Nucleic Acids Res 39, W316–W322.

FEBS Journal 280 (2013) 5237–5257 ª 2013 FEBS 5257