Oncogene (2014) 33, 1190–1197 & 2014 Macmillan Publishers Limited All rights reserved 0950-9232/14 www.nature.com/onc

ORIGINAL ARTICLE PELO negatively regulates HER receptor signalling and metastasis

K Pedersen1, F Canals1, A Prat2, J Tabernero3 and J Arribas1,4,5

The HER family is composed of four receptor tyrosine kinases, which are frequently deregulated in several types of cancer. Activated HER receptors initiate intracellular signalling pathways by attracting to the plasma membrane a plethora of adaptor and signalling molecules. Although there are more than a dozen HER-interacting proteins that regulate signal transduction and have been extensively studied, recent proteomic studies have shown the existence of many novel but largely uncharacterized factors that may bind HER receptors. In this report, we describe a cell-based identification of several new HER2-binding proteins, including HAX1, YWHAZ, PELO and ACP1. Analysis of these factors showed that one of them, PELO, binds to active HER2 and epidermal growth factor receptor and thereby attenuates phosphatidylinositol 3-kinase (PI3K)/AKT signalling, likely through regulation of the recruitment of p85-PI3K to activated receptor. Functional characterization of PELO showed that it negatively regulates cell migration and metastasis in vivo. These results reveal that PELO is a novel regulator of HER-signalling and therefore is likely to have a role in inhibiting tumour progression and invasion.

Oncogene (2014) 33, 1190–1197; doi:10.1038/onc.2013.35; published online 25 February 2013 Keywords: breast cancer; EGFR; metastasis; p95HER2; PELO; PI3K

INTRODUCTION anti-cancer drugs are directed against components of these 6 The epidermal growth factor receptor (EGFR, also known as HER1 pathways. or ERBB1) is the prototype of a receptor tyrosine kinase family that Different in vitro studies have unveiled the existence of many 7–9 comprises three additional members: HER2 (neu, ERBB2), HER3 novel partners of the HER receptors. The interaction of most of and HER4 (ERBB3 and ERBB4). HER receptors react to the binding these proteins with HER receptors in vivo remains to be confirmed of epidermal growth factor (EGF)-like ligands by forming homo- or and their functional relevance are currently unknown. Prompted heterodimers. Despite a strong overall similarity, each of the by these studies, we conducted a proteomic analysis to identify receptors has its own unique functional properties. Remarkably, new HER2-binding proteins using as a bait the HER2 fragment HER2 does not bind any known ligands, instead the structure of its known as 100–115-kDa p95HER2 (also known as 611-CTF). This extracellular domain is constitutively ready for heterodimerization truncated form of HER2, hereafter referred to simply as p95HER2, with a ligand-bound family member.1 The dimerization results in lacks most of the extracellular domain but is constitutively active juxtaposition and subsequent activation of the intracellular kinase because of its ability to form homodimers maintained by 10 domains. Autophosphorylation of certain tyrosine residues in the disulphide bonds. carboxyl-terminal domain of the activated receptors triggers the PELO, one of the novel p95HER2-interacting proteins we recruitment of proteins bearing phosphotyrosine-binding identified and validated, is an evolutionary conserved and domains, that is, SRC-homology-2 or phosphotyrosine-binding pleiotropic protein. In Drosophila melanogaster males with domains.2 Some of these proteins, such as SHC1 or GRB2, mutations in pelota, the gene that encodes for PELO, germ cells have been extensively characterized and shown to contribute to undergo normal mitoses. However, they are not able to complete the initiation of intracellular signalling pathways, which in turn the meiotic division as they remain arrested in late prophase. In regulate different gene expression programs.3 Other proteins, contrast, in mutant D. melanogaster females, the mitotic division is such as the negative regulator of HER-signalling ERRFI1 (also affected during oogenesis. In addition to these germline known as MIG6 or RALT), interact with HER receptors defects, the eyes of pelota mutant flies display impaired independently of a phosphotyrosine-binding motifs.4 development.11 In yeast, the accumulation of free ribosomes HER receptors often have a causal role in malignant progres- and a decrease in the number of polyribosomes in dom34 (the sion. They have been found overactivated in a variety of ortholog of pelota) mutants, indicates that PELO participates in the solid tumours through different mechanisms, including over- regulation of mRNA translation.12 Mice with pelo knockout fail to expression or acquisition of activating mutations.5 In fact, develop past day 7.5 of embryogenesis due to defects in cell HER receptors are the targets of different drugs, including proliferation.13 Interestingly, heterozygous pelo knockout mice tyrosine kinase inhibitors and monoclonal antibodies, currently showed an increase in the incidence of benign tumours, however, used to treat different cancers. Furthermore, signalling pathways the link between PELO and cancer progression remained initiated by the activated HER receptors, such as the RAS/MEK, and unexplored.13 the phosphatidylinositol 3-kinase (PI3K)/AKT pathways are Here we show that PELO binds to active HER2 and EGFR in frequently overactivated in human tumours, and numerous new different breast cancer cell lines. The binding of PELO to these

1Preclinical Research Program, Vall d’Hebron Institute of Oncology (VHIO), Barcelona, Spain; 2Translational Genomics Research Program, VHIO, Barcelona, Spain; 3Clinical Research Program, VHIO, Barcelona, Spain; 4Department of Biochemistry and Molecular Biology, Universitat Autonoma de Barcelona, Campus de la UAB, Bellaterra, Spain and 5Institucio´ Catalana de Recerca i Estudis Avanc¸ats (ICREA), Barcelona, Spain. Correspondence: Dr K Pedersen or Dr J Arribas, Preclinical Research Program, Vall d’Hebron Institute of Oncology, Psg. Vall d’Hebron 119-129, 08035 Barcelona, Spain. E-mail: [email protected] (KPJ) or [email protected] (JA) Received 13 September 2012; revised 14 December 2012; accepted 21 December 2012; published online 25 February 2013 PELO, HER2 and EGFR signalling and metastasis K Pedersen et al 1191 receptors exerts an inhibitory effect as its knockdown resulted in approach, this group of 134 proteins included several factors whose upregulation of the PI3K/AKT pathway activity and in increased interaction with HER2 has been extensively characterized, such as cell migration. Consistently with these results in vitro, PELO SHC1, GRB2 and ERRFI15 (Figure 1c, Supplementary Figure S1 and knockdown cells are more metastatic in vivo. Our results show that Supplementary Table S1). PELO is a novel HER receptor-binding protein that participates in As p95HER2 is constitutively active, we expected that some of the regulation of cell migration and metastasis. the proteins with a H/L ratio 41 would bind specifically to active HER2. To test this and to complement the proteomic approach shown in Figure 1b and c, we performed a similar analysis in cells expressing p95HER2-FLAG treated with or without Lapatinib, a RESULTS small molecule inhibitor that targets the tyrosine kinase activity of Proteomic identification of novel p95HER2-interacting proteins HER2 (Supplementary Figure S1). Confirming our expectation, 91 To identify novel HER2-interacting proteins we used MCF7 Tet-Off of the 134 factors identified comparing control cells with cells cells stably transfected with p95HER2, a truncated and constitu- expressing p95HER2 (Figure 1b and c) were also identified by tively active form of HER2, tagged with a C-terminal FLAG epitope. comparing cells expressing p95HER2 treated with or without Note that p95HER2 migrates as two bands, which correspond to lapatinib (Supplementary Figure S1 and Supplementary Table S2). the immature intracellular precursor and the fully glycosylated The combined results of the two screenings resulted in a list of form located at the cell surface.10 As expected, expression of 261 factors (91 were identified in both screenings whereas 43 and p95HER2-FLAG resulted in the acute activation of the MEK/ERK1,2 127 were identified only in the À / þ doxycycline and À / þ pathway (Figure 1a). lapatinib screenings, respectively). Functional annotation of the Following a Stable Isotope Labelling with Amino Acids in Cell factors identified showed that 28 of them could be potentially Culture (SILAC) approach, and to subsequently trace proteins from involved in intracellular signal transduction (Supplementary cells treated with or without doxycycline (that is, control cells or Figure S1C). cells expressing p95HER2-FLAG), we introduced in the culture media arginine and lysine containing light or heavy isotopes (Figure 1b). After 24 h with or without p95HER2-FLAG expression Validation of selected candidate p95HER2-interacting proteins cell lysates were subjected to anti-FLAG chromatography. Purified We found specific antibodies against 13 of the 28 candidate proteins were analysed by mass spectrometry (see Materials and proteins putatively involved in signal transduction. To validate the Methods) and for every protein identified the ratio of enrichment interaction of these 13 factors to p95HER2, using GRB2 as a resulting from p95HER2-FLAG pull-down was calculated (Figure 1c). positive control, we analysed their levels in p95HER2-FLAG The levels of the majority of proteins did not vary upon expression immunoprecipitates by western blot. As shown in Figure 2a, the of p95HER2-FLAG, indicating that they bound to the beads or the levels of all proteins analysed were enriched in anti-FLAG FLAG antibody and not to the active HER2 fragment. However, the immunoprecipitates from p95HER2-FLAG-expressing cells. heavy/light (H/L) ratio 41 of a group of proteins indicated that they Furthermore, the interaction of several of them with p95HER2 interacted with p95HER2-FLAG (Figure 1c). Validating the proteomic was clearly impaired by Lapatinib. As a result of this analysis, we

Figure 1. Proteomic identification of novel HER2-interacting proteins. (a) MCF7 Tet-Off/p95HER2-FLAG cells were cultured in the presence or absence of doxycycline. At the specified time points, cells were lysed and lysates analysed by western blot with the indicated antibodies. (b) Schematic showing the SILAC protocol used. MCF7 Tet-Off/p95HER2-FLAG cells were cultivated with either ‘light’ or ‘heavy’ arginine and lysine. Cells labelled with light (L) or heavy (H) isotopes for six generations were treated for 24 h with or without doxycycline, respectively. Isotopically labelled proteins were subjected to FLAG chromatography to purify p95HER2-FLAG along with associated proteins. After elution with FLAG peptides, enriched proteins were fractionated through a one-dimension SDS–polyacrylamide gel electrophoresis, digested with trypsin, and analysed by liquid chromatography and mass spectrometry. (c) Average peptide ratios heavy/light of the proteins identified distributed in the indicated intervals and plotted against number of proteins in each interval. The average ratios of SHC1, GRB2 and ERRF11 are shown.

& 2014 Macmillan Publishers Limited Oncogene (2014) 1190 – 1197 PELO, HER2 and EGFR signalling and metastasis K Pedersen et al 1192 interacted with both HAX1 and PELO in MDA-MB-231 cells. These results demonstrate that HAX1 and PELO bind preferentially to active HER2 in different cell lines and that these factors also interact with the EGFR.

PELO attenuates activation of the PI3K/AKT pathway by inhibiting the interaction between EGFR and p85-PI3K To functionally characterize the novel HER2- and EGFR-interacting proteins, we analysed the effect of knocking down HAX1 or PELO from MDA-MB-231 cells treated with EGF. Although the knock- down of these factors did not have an effect on activation of the MEK/ERK1,2 pathway (Supplementary Figure S3), reduction in the levels of PELO, but not those of HAX1, resulted in an increase in the activation of the PI3K/AKT pathway as judged by the analysis of the levels of P-AKT levels (Figure 4a–c). As a control, we showed that the knockdown of SHC1 led to a decreased activation of the PI3K/AKT pathway. We confirmed the relevance of the effect PELO knockdown had on PI3K/AKT by analysing the levels of P-PRAS40, a substrate of P-AKT15 (Supplementary Figure S4A and B). In addition, knockdown of PELO also led to an increased activation of the mammalian target of rapamycin pathway, measured as an elevated level of P-S6, after stimulation of EGFR (Supplementary Figure S4C and D). Although, within HER receptors, HER3 is considered the main Figure 2. Validation of candidate p95HER2-binding proteins. (a) activator of the PI3K/AKT pathway (reviewed in Baselga and MCF7 Tet-Off/p95HER2-FLAG cells treated with or without doxycy- Swain16), EGFR and HER2 can also activate the PI3K/AKT pathway cline and lapatinib as indicated were lysed and immunoprecipitated by recruiting the p85 regulatory subunit of PI3K (PIK3R1), directly with anti-FLAG antibodies. The lysates and immunoprecipitates or via the adaptor protein Gab1.17,18 As MDA-MB-231 cells do not were analysed by western blot with the indicated antibodies. (b) The express HER3 or HER4,19 we concluded that activation of the blots shown in a were quantified and the results expressed as fold PI3K/AKT pathway induced by EGF in these cells (Figures 4a and b) change relative to the signal from immunoprecipitates prepared from cells treated with doxycycline (that is, cells not expressing is mediated, at least in part, by the recruitment of p85-PI3K to p95HER2). activated EGFR. Supporting this conclusion, the levels of P-AKT were reduced by Iressa, a specific EGFR tyrosine kinase inhibitor, in MDA-MB-231 cells treated with EGF (Supplementary Figure S4E). Furthermore, the knockdown of PELO induced a clear increase in the levels of p85-PI3K that interacted with EGFR (Figure 4d), selected HAX1, YWHAZ, PELO and ACP1 for further characteriza- indicating that activation of the pathway induced by low levels of tion because the enrichment of these proteins in p95HER2-FLAG PELO is mediated by an increase in the recruitment of p85-PI3K to immunoprecipitates was similar to that of GRB2 and because the active EGFR. These results strongly suggest that activation of the treatment with Lapatinib clearly impaired the binding of these PI3K/AKT pathway by EGFR is under the control of PELO, a protein factors to p95HER2 (Figure 2b). HSP90 was not further analysed that interacts with the receptor and negatively regulates the because its interaction with HER2 has been previously recruitment of p85-PI3K. characterized.14 Knockdown of PELO increases cell migration and invasion HAX1 and PELO interact with active full-length HER2 and EGFR Although the PI3K/AKT pathway is known to positively regulate To determine whether the p95HER2-binding proteins identified cell proliferation and survival, the knockdown of PELO led to a also interact with full-length HER2, we used breast cancer cells slight reduction in the rate of cell proliferation in vitro (Figure 5a). transduced with HER2, as well as BT474 (a breast cancer cell line However, it concomitantly resulted in an increased cell migration that naturally overexpresses the tyrosine kinase receptor). All the through uncoated membranes in a Boyden chamber (Figures 5b factors analysed interacted with HER2 expressed in MDA-MB-231 and c). Twelve hours after seeding cells on top of the filter, the or MCF7 cells (Figures 3a and b). The analysis with BT474 cells numbers of control and PELO knockdown cells attached at the confirmed that the binding of HAX1 and PELO to HER2 was bottom of the filter were similar (Figure 5b). However, the number impaired upon inhibition of the tyrosine kinase of the latter. The of PELO knockdown cells that, after migrating through the filter, levels of these factors in HER2 immunoprecipitates were reduced detached and accumulated at the bottom of the well was by Lapatinib treatment by 61 and 56%, a reduction similar to that significantly higher than the number found with control cells of GRB2 (Figure 3c). In contrast, the treatment with Lapatinib only (Figures 5b and c). A similar reduced adhesion has been observed reduced moderately the levels of YWHAZ (36%) and ACP1 (12%). in cells that shift from mesenchymal-to-amoeboid migration (see, Most of the factors known to bind HER2 also interact with other for example, Oppel et al.20). Consistent with its lack of effect on the HER receptors.8 Therefore, we tested the interaction of HAX1 and intracellular signalling pathways initiated by HER receptors, the PELO with EGFR. To do so, first we analysed the expression levels knockdown of HAX1 did not have an effect on cell migration of HAX1 and PELO in different breast cancer cell lines. All the cell (Figure 5b and c). lines analysed expressed detectable levels of both factors, albeit at To confirm these results and to generate cells with stable different levels. We chose MDA-MB-231 cells because they can be knockdown of PELO, we transduced MDA-MB-231 cells with used in a variety of functional assays (see below) and they vectors expressing two different shRNAs targeting PELO. Selected expresses readily detectable levels of HAX1 and PELO, and high clones of the two shRNAs displayed levels of PELO around 55 and levels of the EGFR (Supplementary Figure S2A). As shown in 31% relative to cells expressing a control shRNA (Supplementary Figure 3d–f and Supplementary Figure S2B, EGFR efficiently Figure S5A–D). The stably reduced levels of PELO did not

Oncogene (2014) 1190 – 1197 & 2014 Macmillan Publishers Limited PELO, HER2 and EGFR signalling and metastasis K Pedersen et al 1193

Figure 3. HAX1 and PELO interact with active full-length HER2 and EGFR (a) MDA-MB-231 cells transduced with HER2 or HER2-FLAG were lysed and immunoprecipitated with anti-FLAG antibodies. The lysates and immunoprecipitates were analysed with the indicated antibodies by western blot. (b) MCF7 Tet-Off/HER2-FLAG cells cultured without doxycycline and treated with or without lapatinib were lysed and analysed as in a.(c) BT474 cells treated with or without lapatinib were lysed and immunoprecipitated with Herceptin. The lysates and immunoprecipitates were analysed with the indicated antibodies by western blot. (d) MDA-MB-231 cells transduced with EGFR or EGFR-FLAG were lysed and immunoprecipitated with anti-FLAG. The lysates and immunoprecipitates were analysed with the indicated antibodies by western blot. (e, f) MDA-MB-231 cells transduced with HAX1, HAX1-Myc (E), PELO or PELO-FLAG (F) were lysed and immunoprecipitated with anti-Myc or anti-FLAG antibodies, respectively. The lysates and immunoprecipitates were analysed by western blot with the indicated antibodies. significantly reduce cell proliferation (Figure 5d), but augmented injected with cells expressing the control shRNA (Figure 6b, upper cell migration through uncoated membranes (Figure 5e, left), lane). In contrast, metastatic growth was evident in three out of confirming the results in Figure 5b and c. Treatment with the PI3K eight animals (38%) injected with PELO knockdown cells inhibitor GDC-0941 inhibited the migration of PELO knockdown (Figure 6b, middle lane). When the animals were killed, ex vivo cells to levels similar to those of control cells (Figure 5e, right, and analysis of luminescence confirmed the metastatic growth and Supplementary Figure S4F), indicating that the activation of the showed that the metastasis detected in vivo were located at the PI3K/AKT pathway induced by the low levels of PELO is axillary lymph nodes and/or the lungs (Figure 6b, lower lane; responsible for the increase in cell migration. The effect of PELO Supplementary Figure S7). knockdown on cell migration was also evident when using filters Collectively, the results presented here show that PELO interacts covered with matrigel (Figure 5f), suggesting that downmodula- with HER2 and EGFR, and that it exerts an inhibitory effect on the tion of PELO enhances cell invasiveness. PI3K/AKT pathway by negatively regulating recruitment of the p85 regulatory subunit of PI3K. The upregulation of the PI3K/AKT pathway induced by lowering the levels of PELO does not Low levels of PELO result in a more metastatic phenotype significantly affect cell proliferation but increases cell migration Cell migration is a prerequisite for tumour invasion and metastasis. and invasion. These effects are important in vivo, as the To determine whether PELO exerts an inhibitory effect on knockdown of PELO resulted in increased metastasis in mice. This metastasis, we transduced cells knocked down for PELO with a suggests that PELO is likely to be relevant in preventing malignant vector expressing luciferase, a widely used reporter to monitor progression in human cancer. metastasis in vivo (see, for example, Kang et al.21), and injected the resulting cells, or corresponding control cells, orthotopically into nude mice. DISCUSSION In agreement with its lack of effect on cell proliferation Recent high-throughput in vitro approaches have analysed the (Figure 5d), growth of the primary tumours was not affected by interactome of the intracellular domain of activated HER the stable knockdown of PELO (Figure 6a; Supplementary receptors.7–9 Despite the limitations of in vitro models, such as Figure S6). When the tumours attained a volume of B400 mm3, aberrant folding of recombinant proteins and peptides or lack they were surgically removed. Two months after primary tumour of physiological conditions for protein–protein interactions, resection we did not detect any sign of metastatic growth in mice these reports showed the existence of many uncharacterized

& 2014 Macmillan Publishers Limited Oncogene (2014) 1190 – 1197 PELO, HER2 and EGFR signalling and metastasis K Pedersen et al 1194

Figure 4. Knockdown of PELO results in overactivation of the PI3K/AKT pathway. (a) MDA-MB-231 cells treated with control siRNAs or specific siRNAs targeting HAX1, PELO or SHC were cultured in media without serum for 12 h, treated with EGF for different periods of time and lysed. The lysates were analysed by western blot with the indicated antibodies. (b) The results of three independent experiments performed as in a were quantified and expressed as averages ±s.d. P-value was obtained by two-tailed Student’s t-test, *Po0.05. (c) The cell lysates corresponding to time 0 from a were analysed by western blot with the indicated antibodies. (d) MDA-MB-231 cells transduced with EGFR- FLAG were treated with the EGFR inhibitor Iressa, control siRNAs and/or siRNAs targeting PELO, as indicated. After ON starvation, cells were stimulated with EGF for 30 min, lysed and immunoprecipitated with anti-FLAG antibodies. The lysates and immunoprecipitates were analysed by western blot with an antibody against p85-PI3K.

HER receptor-interacting proteins. Using a cell-based model we the PI3K/AKT by EGFR or HER2, we chose as an experimental have confirmed the complexity of the interactome of HER system the MDA-MB-231 cell line, which lack HER3 and HER4.19 In receptors; in addition to extensively characterized factors known these cells, we showed that the PI3K/AKT pathway is activated by to bind HER receptors, including ERFFI1, GRB2, SHC1, HSP90 or the recruitment of the p85 regulatory subunit and that the PIK3R1, we identified several new partners of HER2. contribution of activated RAS is limited, if any. This conclusion is Some of the factors identified (HAX1, YWHAZ, PELO and ACP1) based on the fact that although the knockdown of PELO activated preferentially bound to active HER2. However, none of them the PI3K/AKT pathway, it had no effect on the levels of phospho- contain SRC-homology-2 or phosphotyrosine-binding domains, ERK1,2 (Supplementary Figure S3), a downstream target of the indicating that if they interact directly with HER2, they do not bind RAS/MEK pathway. Furthermore, although the knockdown of to phosphotyrosine residues. Alternatively, these factors may PELO did not affect the RAS/MEK pathway, it did have a clear contain unknown phosphotyrosine-binding domains. The second effect on the recruitment of p85-PI3K and activation of the PI3K/ possibility is feasible as novel phosphotyrosine-binding domains AKT pathway (Figure 4). are still being discovered and characterized.22 Another possibility Our demonstration of a novel receptor tyrosine kinase/PI3K is that binding to phosphotyrosines is indirect, through a protein regulatory function of PELO does not exclude that the previously containing a SRC-homology-2 or phosphotyrosine-binding described functions in regulation of translation12 and/or cell domain. Future characterization of the binding of these factors division11,13 could be equally important in breast cancer cells. to HER2 and EGFR, using recombinant purified proteins, will help Indeed, very efficient knockdown of PELO with siRNAs (Figure 4c) to distinguish between these possibilities. did lead to a reduction in cell proliferation (Figure 5a), which we Although we validated the binding of the four factors did not expect based on the increased activity of the PI3K/AKT/ mentioned to ectopically expressed HER2 in different cell lines, mammalian target of rapamycin pathway in those cells (Figure 4a only HAX1 and PELO bound preferentially to active endogenous and Supplementary Figure 4A–D). HER2 and EGFR. Curiously, it has been shown recently that PELO Two lines of evidence linking low levels of PELO to cancer interacts with HAX1.23 However, the interaction of PELO with HER2 progressions have been published. Heterozygous pelo knockout or EGFR is not likely to occur via HAX1 as the knockdown of the mice show an increase in the incidence of benign tumours.13 latter has no effect on PI3K/AKT signalling whereas the Presumably, these mice expressed levels of PELO similar to those knockdown of PELO activates this pathway. However, we cannot in our knockdown cell lines (Supplementary Figure S5 and S7); rule out that the interaction of HAX1 with the receptor is therefore, the results with knockout mice, as well as ours, point to established via PELO. a tumour-suppressor effect of PELO. In support of this conclusion, Receptor tyrosine kinases activate the PI3K/AKT pathway a heterozygous non-sense mutation, which introduces a stop directly, via the recruitment of the p85 regulatory subunit unit codon after amino acid 26 and generates a protein product most of PI3K, or indirectly, via RAS; the activated small GTPase directly likely inactive, has been described recently in an ovary serous binds and activates PI3K (for a recent review see Engelman6). As carcinoma.24 Interestingly, an inspection of the PELO mRNA levels HER3 and, to a minor extent, HER4 could mask the activation of in two independent publically available genomic data sets

Oncogene (2014) 1190 – 1197 & 2014 Macmillan Publishers Limited PELO, HER2 and EGFR signalling and metastasis K Pedersen et al 1195

Figure 5. Knockdown of PELO promotes cell migration and invasion. (a) The relative number of MDA-MB-231 cells treated with control siRNAs or siRNAs targeting HAX1 or PELO was estimated using the WST1 assay at the specified time points. The bars represent averages ±s.d. of three independent experiments. P-value was obtained by two-tailed Student’s t-test, **Po0.01. (b, c) MDA-MB-231 cells treated with control siRNA or siRNAs targeting HAX1 or PELO were seeded in trans-well plates. After 12 h, cells at the bottom of the filter were fixed, stained and counted (filter). The number of cells that detached from the filter and accumulated in the well was also counted (well). Representative fields are shown in b. The total number of cells migrated in three independent experiments were analysed and bars representing averages ±s.d. are shown in c. P-value was obtained by two-tailed Student’s t-test, **Po0.01. (d) MDA-MB-231 cells were transduced with a lenti-viral vector encoding a control shRNA or one of two independent shRNAs targeting PELO. Stable clones expressing the specified shRNAs were selected and cell proliferation analysed as in a.(e) Left, migration of the same cells as in d was analysed in Boyden chambers as in c. Right, migration of cells expressing one of the shRNAs targeting PELO with and without treatment with the PI3K inhibitor GDC-0941. (f) Migration through matrigel-coated filters of the same cells as in d was analysed in Boyden chambers as in c, except the allowed time of migration was 36 h. The bars represent averages ±s.d. of the results from three independent experiments. P-values were obtained by two-tailed Student’s t-test, *Po0.05, **Po0.01.

Figure 6. Knockdown of PELO results in a more metastatic phenotype. (a) 1.5 Â 106 MDA-MB-231 cells expressing firefly luciferase and a control shRNA or a shRNA targeting PELO were injected into the mammary fat pad of nude mice (n ¼ 9 and n ¼ 8 for control shRNA and shRNA targeting PELO, respectively). Tumour volume was monitored at the indicated time points and the results were expressed as averages±s.d. (b) Two months after removal of the primary tumour, metastatic growth was detected by in vivo imaging of total photon flux. Next, mice were killed and luminescence was analysed in axillary lymph nodes and lungs. Identical settings and scale (shown) were used in image preparation for all panels. revealed that it is particularly underexpressed in the aggressive the basal-like subtype had frequent deletion of the pelo gene basal-like breast cancer (Supplementary Figure S8A and B), a (Supplementary Figure S8D). subtype that frequently exhibits overactivation of the EGFR.25 In conclusion, the results presented here unveil the existence of Although a correlation analysis including all breast cancer a novel partner of activated HER2 and EGFR with the ability to subtypes displayed an inverse correlation between PELO levels regulate the PI3K/AKT pathway and as a result of that cell and PI3K/AKT activity (Supplementary Figure S8C), only samples of migration and metastasis.

& 2014 Macmillan Publishers Limited Oncogene (2014) 1190 – 1197 PELO, HER2 and EGFR signalling and metastasis K Pedersen et al 1196 MATERIALS AND METHODS material was not combined but concentrated and fractionated individually. Cell lines In the case of Myc-tag HAX1, we used 80 ml anti-Myc-coupled beads (Cell Signaling) for 4 mg total protein per sample and enriched proteins were The cell lines MCF10A, MDA453, MDA-MB-231, BT474, MCF7, T47D and eluted with 1 SDS loading buffer. To pull-down endogenous HER2 from SkBr3 were purchased from ATCC (Manassas, VA, USA). Unless otherwise  mM stated MDA-MB-231, BT474, MCF7, T47D and SkBr3 were grown at 37 1Cin BT474 cells with or without 24-h pretreatment with 0.25 Lapatinib, we 1 presence of 5% CO and in DMEM/F-12 (Invitrogen, Paissley, UK) incubated 80 mg total protein with 3.2 mg Herceptin (Roche) at 4 C for 2 h, 2 added 3.2 ml Protein A beads (GE Healthcare, Uppsala, Sweden) and supplemented with 8% fetal bovine serum (FBS) and 4 mML-glutamine. 1 MDA453 was propagated without CO in Leibovitz’s L-15 (Invitrogen) incubated the sample an additional 1 h at 4 C. Due to the amount of 2 beads used they were washed by spin precipitation, but the 1  SDS supplemented with 8% FBS and 4 mML-glutamine. The MCF10A cells were loading buffer elutions were passed through a 0.22-mm PVDF Centrifugal maintained with 5% CO in DMEM/F-12 supplemented with 8% FBS, 4 mM 2 Filter Unit. Likewise, the endogenous EGFR from MDA-MB-231 cells was L-glutamine, 9 mg/ml insulin (Sigma, Steinheim, Germany), 0.5 mg/ml hydrocortisone and 20 ng/ml EGF. The stable clones MCF7 Tet-Off/p95- immunoprecipitated with Cetuximab (Bristol-Myers Squibb, New York, NY, HER2-FLAG and MCF7 Dox-Off/HER2-FLAG were selected as describe for USA) using IgG1 (antibodies-online.com, Aachen, Germany) as a negative 611-CTF and HER2, respectively, in Pedersen et al.,10 except a sequence control. In samples where a kinase inhibitor had been used in the culture encoding the FLAG epitope was inserted before the stop codon at the 30 medium it was also included in the IP lysis buffer. end of the coding regions. Transduction SILAC-based screening Expression of FLUC, HER2 or HER2-FLAG in MDA-MB-231 cells was To follow the proteins from cells with and without expression of p95-HER2- accomplished by cloning the corresponding sequences in the retro-viral FLAG, we used SILAC with components from Invitrogen and according to 26 vector pQCXIH (Clontech, Mountain View, CA, USA), whereas EGFR, EGFR- the principles described in Blagoev et al. MCF7 Tet-Off/p95-HER2-FLAG FLAG, PELO, PELO-FLAG, HAX1 and HAX1-Myc were expressed from the cells were grown with 1 mg/ml doxycycline in SILAC DMEM medium lenti-viral vector pLEX (Open Biosystems, Thermo Scientific ABgene, supplemented with 8% dialyzed serum, 4 mML-glutamine and the Epsom, UK). Viruses from pQCXIH were prepared with the GP2-293 appropriate normal amino acids, in addition to isotopically distinct packaging cell line, the envelope encoding plasmid pVSV-G and arginine and lysine (12C14N ‘light’ arginine and 12C ‘light’ lysine or Lipofectamine 2000 (Invitrogen). Lenti viruses were prepared with 13C15N ‘heavy’ arginine and 13C ‘heavy’ lysine) at 45 mg/ml and 90 mg/ml, HEK293T and Trans-Lenti Packaging kit from Open Biosystems. Virus respectively, for 6 cell-doublings. Cells labelled with light or heavy isotopes containing supernatants were filtered with 0.45-mm PVDF filters (Millipore), were detached with trypsin/EDTA and washed three times in a medium and medium from 1 cm2 packaging cells was used to infect 2 cm2 MDA- with and without doxycycline, respectively, and seeded at 40% confluence. MB-231 cells at B35% confluence. Approximately 90% infection efficiency After 12 h the medium was changed and dishes were incubated for an was verified by a parallel control expressing GFP around 3 days after addition 12 h with or without doxycycline. Light and heavy total protein transduction. lysates were prepared, 22 ml each of 2 mg/ml, by lysing the cells in MDA-MB-231 cells expressing B55 and 31% of the levels of PELO found immunoprecipitation (IP) lysis buffer: 1% NP-40, 150 mM NaCl, 50 mM Tris in the parental cells were selected with 0.8 mg/ml puromycin starting 3 pH 7.4, 5 mM b-glycerophosphate, 5 mM NaF, 1 mM Na3VO4 and Complete days after transduction with the lenti-viral plasmids (Open Biosystems) protease inhibitor cocktail (Roche, Mannheim, Germany). The lysates were pGIPZ-shRNAmir-PELO clone V3LHS_328599 and V3LHS_328597. Cells incubated with 0.25 ml anti-FLAG M2 agarose beads with gentle tilting at transduced with the pGIPZ-Non-Silencing-Control expressed the same 1 4 C for 3 h. After a brief centrifugation, 90% of the supernatants were levels of PELO as parental cells. aspirated and the beads were transferred to Ultrafree-CL Centrifugal Filter Units with 0.22 mm polyvinylidene difluoride (PVDF) membranes (Millipore, Billerica, MA, USA) and washed with 3  2 ml lysis buffer. Bound proteins were eluted with 1 ml 100 mg/ml 3  FLAG peptide by incubation at room siRNA knockdown temperature for 1 h. Next the light and heavy lysates were combined and The levels of HAX1, PELO or SHC1 were transiently knocked down concentrated 10 times with 3 kDa Amicon Ultra-0.5-ml Centrifugal filters by Lipofectamine RNAiMAX reverse transfection of the corresponding (Millipore). Half of the concentrated sample was fractionated in one lane of ON-TARGETplus siRNAs (Dharmacon, Chicago, IL, USA). The sequences a precast NuPAGE 4–12% Bis-Tris gel 1.5 mm 10 well (Invitrogen). The one- were the following: PELO (50-GAAAUAAUCUCCACGUACU-30,50-UUAAAUG dimensional gel lane was cut into 20 horizontal slices, and each slice was AUUGCCGUACAA-30,50-GUGUGGUACUGGAGCGCAU-30 and 50-GCGUGGAG subjected to in-gel tryptic digestion and liquid chromatography–mass GCCAUCGACUU)-30; HAX1 (50-GGAUACGUUUCCACGAUAA-30,50-UACAGU spectrometry analysis as described in Garcia-Castillo et al.27 For protein AACCCGACACGAA-30,50-GGACAGACACUUCGGGACU-30 and 50-AAUAGCAU quantitation, H/L ratios were calculated averaging the measured H/L ratios CUUCAGCGAUA-30); SHC1 (50-GACAAUCACUUGCCCAUCA-30,50-GAGUUGC for the observed peptides, after discarding outliers. For selected proteins of GCUUCAAACAAU-30,50-CACGGGAGCUUUGUCAAUA-30 and 50-GACUAAG interest, quantitation of data obtained from the automated WARP-LC GAUCACCGCUUU-30). As a negative control cells were transfected with the analysis was manually reviewed. ON-TARGETplus Non-Targeting Pool siRNAs (ref 77D–001810–10–20). Cells were lysed or seeded for functional analyses 4 days after transfection. Western blot When mentioned, cells were washed and kept in medium over-night (ON) Functional assays without FBS, followed by stimulation with 2 nM EGF. Where indicated, EGFR, HER2 or PI3K was inactivated by ON treatment with 1 mM Iressa (AstraZeneca, Cell proliferation was followed by measurement of WST1 (Roche) London, UK), 2 mM Lapatinib (GlaxoSmithKline, London, UK) or 0.5 mM GDC- conversion. Cells were detached with trypsin/EDTA, at time 0 cells (2000/ 0941 (Selleckchem, Munich, Germany). Total protein cell lysates were well) were seeded in a 96-well plate and incubated. prepared with IP lysis buffer including 0.25% deoxycholate. Fractionation of Cell migration and invasion were determined with 24-well format 15 mg total protein per sample was done in NuPAGE 4–12% Bis-Tris gels. For Boyden chambers containing FluoroBlok PET membranes with 8-mm pores blotting we used PVDF membranes (Millipore, Carrigtwohill, Ireland). without or with precoated matrigel (BD biosciences). Wells were filled with Antibodies against the following proteins were used for probing: AKT, medium containing 0.5% FBS and 2 nM EGF. ON FBS-starved cells were CTNNA1, EEF1A1, EGFR, HSP90, p85-PI3K, P-AKT, P-ERK1/2, P-HER2, P-PRAS40, detached with 10 mM EDTA, washed and resuspended in medium without P-S6 and YWHAZ (Cell Signaling, Leiden, the Netherlands); ACP1, DSTN, FBS. At time 0 cells (20 000/insert) were seeded on top of membranes and NME1, PELO, PFN1, RAB13, TBL2 (Sigma, St Louis, MO, USA); GRB2, SHC1 (BD an identical volume of cell suspension was seeded in a standard 24-well Bioscience, San Jose, CA, USA); GAPDH (Trevigen, Gaithersburg, MD, USA); plate as a control of cell number. The assay was terminated by fixing GNAS (AbCam, Cambridge, UK); HAX1 (R&D systems, Minneapolis, MN, USA); membranes with cells in 4% formaldehyde for 15 min at room temperature HER2 (BioGenex, San Ramon, CA, USA). and staining with Green CMFDA CellTracker (Invitrogen). Where shown, the PI3K inhibitor GDC-0941 was included in the well (0.5 mM). For each experiment and condition, migration across the membranes was Immunoprecipitation quantified for eight independent inserts by acquiring an image of the For Immunoprecipitation of receptors with FLAG-tag, we used the same bottom of the membrane and one of the well both at the centre with a procedure as described for the SILAC experiment, except the eluted  10 objective and manual counting of cell numbers.

Oncogene (2014) 1190 – 1197 & 2014 Macmillan Publishers Limited PELO, HER2 and EGFR signalling and metastasis K Pedersen et al 1197 Mouse model of breast cancer metastasis 10 Pedersen K, Angelini PD, Laos S, Bach-Faig A, Cunningham MP, Ferrer-Ramon C Mice were maintained and treated in accordance with institutional et al. A naturally occurring HER2 carboxy-terminal fragment promotes mammary guidelines of Vall d’Hebron University Hospital Care and Use Committee. tumor growth and metastasis. Mol Cell Biol 2009; 29: 3319–3331. 1.5  106 cells were injected into the mammary fat pad of gland 11 Eberhart CG, Wasserman SA. The pelota locus encodes a protein required for number 4 of 7-week-old female BALB/c nude mice purchased from meiotic cell division: an analysis of G2/M arrest in Drosophila spermatogenesis. Charles Rivers Laboratories (L’Arbresle Cedex, France). Tumour xenografts Development 1995; 121: 3477–3486. were measured with calliper every 7 days, and tumour volume was 12 Davis L, Engebrecht J. Yeast dom34 mutants are defective in multiple develop- calculated using the formula: (length  width2)  (pi/6). At the end of the mental pathways and exhibit decreased levels of polyribosomes. Genetics 1998; experiment, the animals were anaesthetized with a 1.5% isoflurane air 149:45–56. mixture and killed by cervical dislocation. Results are presented as 13 Adham IM, Sallam MA, Steding G, Korabiowska M, Brinck U, Hoyer-Fender S et al. mean±s.d. of tumour volume. Metastatic colonization was monitored Disruption of the pelota gene causes early embryonic lethality and defects in cell by in vivo bioluminescence imaging using the IVIS-200 imaging cycle progression. Mol Cell Biol 2003; 23: 1470–1476. system from Xenogen (PerkinElmer, Waltham, MA, USA) as described in 14 Zhou P, Fernandes N, Dodge IL, Reddi AL, Rao N, Safran H et al. ErbB2 Minn et al.28 degradation mediated by the co-chaperone protein CHIP. J Biol Chem 2003; 278: 13829–13837. 15 Rhodes N, Heerding DA, Duckett DR, Eberwein DJ, Knick VB, Lansing TJ et al. Genomic bioinformatics analyses Characterization of an Akt kinase inhibitor with potent pharmacodynamic and Gene expression microarray data, DNA copy number data, and reverse- antitumor activity. Cancer Res 2008; 68: 2366–2374. phase protein array data were obtained from The Cancer Genome Atlas 16 Baselga J, Swain SM. Novel anticancer targets: revisiting ERBB2 and discovering (TCGA) project portal (http://cancergenome.nih.gov/) of the following ERBB3. Nat Rev Cancer 2009; 9: 463–475. tumour types: breast cancer,29 lung squamous cell carcinoma,30 ovarian 17 Rodrigues GA, Falasca M, Zhang Z, Ong SH, Schlessinger J. A novel positive cancer,31 glioblastoma multiforme32 and colorectal cancer.33 Subtypes calls feedback loop mediated by the docking protein Gab1 and phosphatidylinositol were used as provided in each TCGA publication. For gene expression 3-kinase in epidermal growth factor receptor signaling. Mol Cell Biol 2000; 20: microarray analyses, 244K Agilent-based data (Agilent Technologies, Santa 1448–1459. Clara, CA, USA) from all tumour types was filtered by requiring the Lowess 18 Stover DR, Becker M, Liebetanz J, Lydon NB. Src phosphorylation of the epidermal normalized intensity values in both sample and control to be 410. The growth factor receptor at novel sites mediates receptor interaction with Src and normalized log2 ratios (Cy5 sample/Cy3 control) or log2 intensity of probes P85 alpha. J Biol Chem 1995; 270: 15591–15597. mapping to the same gene (Entrez ID as defined by the manufacturer) were 19 Moasser MM, Basso A, Averbuch SD, Rosen N. The tyrosine kinase inhibitor averaged to generate independent expression estimates. In each cohort, ZD1839 (‘Iressa’) inhibits HER2-driven signaling and suppresses the growth of genes were median centred and standardized to zero mean and unit HER2-overexpressing tumor cells. Cancer Res 2001; 61: 7184–7188. variance. To confirm the findings in breast cancer, an independent, and 20 Oppel F, Muller N, Schackert G, Hendruschk S, Martin D, Geiger KD et al. SOX2- publicly available, gene expression microarray data set was evaluated, which RNAi attenuates S-phase entry and induces RhoA-dependent switch to protease- is composed of 337 breast samples (UNC337 data set: GSE18229).34 Subtype independent amoeboid migration in human glioma cells. Mol Cancer 2011; 10: calls were used as provided. Finally, we estimated the activation of the PI3K 137. pathway using the TCGA RRPA data and the formula provided in the TCGA 21 Kang Y, Siegel PM, Shu W, Drobnjak M, Kakonen SM, Cordon-Cardo C et al. A breast cancer report. multigenic program mediating breast cancer metastasis to bone. Cancer Cell 2003; 3: 537–549. 22 Mukherjee M, Chow SY, Yusoff P, Seetharaman J, Ng C, Sinniah S et al. Structure of CONFLICT OF INTEREST a novel phosphotyrosine-binding domain in Hakai that targets E-cadherin. EMBO J 2012; 31: 1308–1319. The authors declare no conflict of interest. 23 Burnicka-Turek O, Kata A, Buyandelger B, Ebermann L, Kramann N, Burfeind P et al. Pelota interacts with HAX1, EIF3G and SRPX and the resulting protein complexes are associated with the actin cytoskeleton. BMC Cell Biol 2010; 11:28. ACKNOWLEDGEMENTS 24 Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian This work was supported by the Instituto de Salud Carlos III (Intrasalud PI081154 and carcinoma. Nature 2011; 474: 609–615. the network of cooperative cancer research (RTICC-RD06/0020/0022)), the Breast 25 Nielsen TO, Hsu FD, Jensen K, Cheang M, Karaca G, Hu Z et al. Immunohisto- Cancer Research Foundation, the Asociacio´n Espan˜ola Contra el Cancer (AECC) and chemical and clinical characterization of the basal-like subtype of invasive breast AVON Foundation. KP was supported by the postdoctoral program from the AECC. carcinoma. Clin Cancer Res 2004; 10: 5367–5374. 26 Blagoev B, Kratchmarova I, Ong SE, Nielsen M, Foster LJ, Mann M. A proteomics strategy to elucidate functional protein-protein interactions applied to EGF signaling. Nat Biotechnol 2003; 21: 315–318. REFERENCES 27 Garcia-Castillo J, Pedersen K, Angelini PD, Bech-Serra JJ, Colome N, Cunningham 1 Citri A, Yarden Y. EGF-ERBB signalling: towards the systems level. Nat Rev Mol Cell MP et al. HER2 carboxyl-terminal fragments regulate cell migration and cortactin Biol 2006; 7: 505–516. phosphorylation. J Biol Chem 2009; 284: 25302–25313. 2 Yaffe MB. Phosphotyrosine-binding domains in signal transduction. Nat Rev Mol 28 Minn AJ, Gupta GP, Siegel PM, Bos PD, Shu W, Giri DD et al. Genes that mediate Cell Biol 2002; 3: 177–186. breast cancer metastasis to lung. Nature 2005; 436: 518–521. 3 Hynes NE, MacDonald G. ErbB receptors and signaling pathways in cancer. Curr 29 TCGA. Comprehensive molecular portraits of human breast tumours. Nature 2012; Opin Cell Biol 2009; 21: 177–184. 490:61–70. 4 Segatto O, Anastasi S, Alema S. Regulation of epidermal growth factor receptor 30 TCGA. Comprehensive genomic characterization of squamous cell lung cancers. signalling by inducible feedback inhibitors. J Cell Sci 2011; 124: 1785–1793. Nature 2012; 489: 519–525. 5 Yarden Y, Pines G. The ERBB network: at last, cancer therapy meets systems 31 TCGA. Integrated genomic analyses of ovarian carcinoma. Nature 2011; 474: biology. Nat Rev Cancer 2012; 12: 553–563. 609–615. 6 Engelman JA. Targeting PI3K signalling in cancer: opportunities, challenges and 32 Verhaak R, Hoadley K, Purdom E, Wang V, Qi Y, Wilkerson MD et al. Integrated limitations. Nat Rev Cancer 2009; 9: 550–562. genomic analysis identifies clinically relevant subtypes of glioblastoma char- 7 Jones RB, Gordus A, Krall JA, MacBeath G. A quantitative protein interaction net- acterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer cell 2010; 19: work for the ErbB receptors using protein microarrays. Nature 2006; 439: 168–174. 98–110. 8 Morandell S, Stasyk T, Skvortsov S, Ascher S, Huber LA. Quantitative proteomics 33 TCGA. Comprehensive molecular characterization of human colon and rectal and phosphoproteomics reveal novel insights into complexity and dynamics of cancer. Nature 2012; 487: 330–337. the EGFR signaling network. Proteomics 2008; 8: 4383–4401. 34 Prat A, Parker J, Karginova O, Fan C, Livasy C, Herschkowitz JI et al. Phenotypic and 9 Schulze WX, Deng L, Mann M. Phosphotyrosine interactome of the ErbB-receptor molecular characterization of the claudin-low intrinsic subtype of breast cancer. kinase family. Mol Syst Biol 2005; 1: 2005–0008. Breast Cancer Res 2010; 12: R68.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

& 2014 Macmillan Publishers Limited Oncogene (2014) 1190 – 1197