Loss of the Tumor Suppressor Spinophilin (PPP1R9B) Increases the Cancer Stem Cell Population in Breast Tumors

ORIGINAL ARTICLE Loss of the tumor suppressor spinophilin (PPP1R9B) increases the cancer stem cell population in breast tumors

I Ferrer1, EM Verdugo-Sivianes1, MA Castilla1, R Melendez1, JJ Marin1,2, S Muñoz-Galvan1, JL Lopez-Guerra3, B Vieites4, MJ Ortiz-Gordillo3, JM De León5, JM Praena-Fernandez6, M Perez1, J Palacios7 and A Carnero1,8

The spinophilin (Spn, PPP1R9B) gene is located at 17q21.33, a region frequently associated with microsatellite instability and loss of heterozygosity, especially in breast tumors. Spn is a regulatory subunit of phosphatase1a (PP1), which targets the catalytic subunit to distinct subcellular locations. Spn downregulation reduces PPP1CA activity against the retinoblastoma protein, pRb, thereby maintaining higher levels of phosphorylated pRb. This effect contributes to an increase in the tumorigenic properties of cells in certain contexts. Here, we explored the mechanism of how Spn downregulation contributes to the malignant phenotype and poor prognosis in breast tumors and found an increase in the stemness phenotype. Analysis of human breast tumors showed that Spn mRNA and protein are reduced or lost in 15% of carcinomas, correlating with a worse prognosis, a more aggressive tumor phenotype and triple-negative tumors, whereas luminal tumors showed high Spn levels. Downregulation of Spn by shRNA increased the stemness properties along with the expression of stem-related genes (Sox2, KLF4, Nanog and OCT4), whereas ectopic overexpression of Spn cDNA reduced these properties. Breast tumor stem cells appeared to have low levels of Spn mRNA, and Spn loss correlated with increased stem-like cell appearance in breast tumors as indicated by an increase in CD44+/CD24- cells. A reduction of the levels of PPP1CA mimicked the cancer stem-like cell phenotype of Spn downregulation, suggesting that the mechanism of Spn involves PP1a. These increased cancer stem cell-like properties with reduced Spn might account for the malignant phenotype observed in Spn-loss tumors and may contribute to a worse patient prognosis.

Oncogene (2016) 35, 2777–2788; doi:10.1038/onc.2015.341; published online 21 September 2015

INTRODUCTION colorectal, gastric, renal and lung carcinomas, as well as in salivary Spinophilin (Spn, also known as neurabin II and PPP1R9B) was first gland carcinosarcoma, an extremely aggressive neoplasm. isolated independently by two laboratories and has been shown In extensive LOH mapping in primary lung carcinoma using 15 17 to interact with protein phosphatase 1 (PP1) and F-actin.1,2 The highly polymorphic markers, the highest LOH value appeared domain structure of Spn suggests that it functions as a with the D17S588 marker, showing a 53% loss. This marker multifunctional protein scaffold that regulates protein–protein is located exactly within the Spn locus. However, the neighboring interactions.3,4 More than 30 partner proteins of Spn have been tumor suppressor genes, such as BRCA1, were not significantly discovered3,5–7 to impinge on both membrane and cytoskeletal affected (6–13% LOH). Deep analysis of human lung tumors functions.3,8 However, the physiological relevance of some showed that the Spn protein is absent in 20% of neoplasias and of these interactions remains to be determined. Spn performs reduced in another 38%. Spn reduction correlates with malignant important functions in the nervous system where it has been grade and p53 mutations.18 Low Spn levels have been associated implicated in the regulation of spine morphology and density, with a high proliferative grade and poor patient prognosis in synaptic plasticity and neuronal migration.9,10 hepatocarcinoma, head and neck cancer and advanced stages of – The Spn gene is located at 17q21.33, a cytogenetic area colorectal carcinoma.19 21 frequently associated with microsatellite instability and loss At a functional level, Spn regulates PP1 activity,22 whereas loss of heterozygosity (LOH). Most of the studies on the 17q21 locus of Spn reduces PPP1CA activity on its target pRb,23 thereby have focused on BRCA1, which exhibits a variable frequency in maintaining higher levels of phosphorylated pRb.24 In the absence LOH depending on the type and stage of the tumor.4 However, of p53, reduced levels of Spn increase the tumorigenic properties several studies have suggested the presence of another tumor of cells.24 Spn-KO mice have a reduced lifespan, an increased suppressor gene in this area that includes the Spn locus,11–16 number of tumors and increased cellular proliferation in some which indicates the importance of this locus independent tissues, such as the mammary ducts. In addition, the combined of BRCA1 in the pathology of cancer, especially in breast tumors loss of Spn and p53 activity in mouse models leads to an increase (Supplementary Table I). LOH in chromosome 17q21.33 has been in mammary carcinomas, which confirms the functional relation- observed in different tumors, including breast, ovarian, prostate, ship between p53 and Spn.25

1Instituto de Biomedicina de Sevilla (IBIS/HUVR/CSIC/Universidad de Sevilla), Seville, Spain; 2Department of Preventive Medicine and Public Health, Seville University, Seville, Spain; 3Department of Radiation Oncology, University Hospital Virgen Del Rocío, Seville, Spain; 4Department of Pathology, University Hospital Virgen Del Rocío, Seville, Spain; 5Unidad Funcional de Patología Mamaria, University Hospital Virgen Del Rocío, Seville, Spain; 6Methodology Unit- Fundación Pública Andaluza para la Gestión de la Investigación en Salud de Sevilla, Virgen del Rocío University Hospital, Seville, Spain; 7Department of Pathology, University Hospital Ramon y Cajal, Madrid, Spain and 8Consejo Superior de Investigaciones Cientificas, Sevilla, Spain. Correspondence: Professor A Carnero, Instituto de Biomedicina de Sevilla (IBIS/HUVR/CSIC/US), Campus HUVR, Edificio IBIS, Avda. Manuel Siurot s/n, Sevilla 41013, Spain. E-mail: [email protected] Received 18 December 2014; revised 6 August 2015; accepted 7 August 2015; published online 21 September 2015 Loss of Spn increases the cancer stem cell population in breast tumors I Ferrer et al 2778 The pRb pathway controls several aspects of stem cell biology, obtained from prophylactic surgery. We found that 80–85% of the including the tight control of self-renewal characteristics of tumors displayed high levels of Spn, similar to the values observed progenitor cells.26,27 It is tempting to speculate that the loss in normal glands (Figures 1a and b and Supplementary Figure S1). of Spn may result in a partial expansion of certain progenitor Another 10–15% of tumors showed intermediate levels of Spn pools. This expansion may explain the preneoplastic abnormalities (Figures 1a and b). In addition, 2.4% of tumors showed no associated with increased cell proliferation found in Spn-null Spn expression (Figures 1a and b). We also observed that loss of 24,25 mice. Activation of a p53-mediated arrest may further limit Spn correlated with a higher histological grade determined at the expansion of these cells. Elimination of this additional barrier may time of surgery (Figure 1c) and a less differentiated phenotype therefore explain the synergistic effect observed in terms of tumor (Figure 1d). formation in Spn-null animals when p53 function is impaired. Collectively, our data suggest that Spn is lost in a small Furthermore, Spn can restrain self renewal of brain tumor-initiating percentage of human breast tumors at different extents and that 28 29 cells and anchorage-independent growth of glioma cell lines. lower or null levels of Spn may correlate with a higher In the present work, we have attempted to understand the Spn histological grade. loss-induced causal effect in cancer, specifically in the breast, given the relevance of its locus, which might explain the relevance of its loss during tumorigenesis. We found that the Spn protein is Loss of Spn correlates with worse prognosis and the triple- reduced or lost in 15% of breast carcinomas, which correlates with negative phenotype in breast tumors a more aggressive tumor phenotype. Furthermore, we have To explore the relevance of Spn loss in breast tumors, observed an inverse correlation between Spn and cancer stem we extracted mRNA from tumor samples from a second cohort cells (CSCs, also called cancer-initiating cells). The enforced of human breast tumors comprising different types of tumors. downregulation of Spn by shRNA induced CSC-like properties in The analysis of the levels of mRNA indicates that Spn levels were breast tumor cells accompanied by an increase in stem factors. especially reduced in ER-negative tumors (Student’s t-test We also observed a significant correlation of Spn loss with triple- Po0.001; Figure 2a) and in triple-negative tumors in contrast to negative breast tumors, whereas luminal tumors maintain higher the luminal type tumors (Student’s t-test Po0.001; Figure 2b). levels of Spn mRNA. Therefore, the increased CSC-like properties These data indicate that reduced Spn levels are more likely induced by the downregulation of Spn might account for the to occur in breast tumors with a worse prognosis. There was no increased malignant phenotype observed in Spn-null breast difference in Spn levels between luminal A or B type tumors tumors and its correlation with a poor prognosis. (Figure 2b). Specific analysis of the population of Her2-positive (ER and PR negative) tumors showed that these tumors exhibited lower levels of Spn mRNA (Figure 2c). Similar data were obtained RESULTS from the analysis of cohort 1 by immunohistochemistry Analysis of Spn in human breast tumors (Supplementary Table S4). To study whether reductions in Spn levels are a common event To confirm the effect of Spn loss on patient prognosis, we in human breast tumors, we generated a tissue array of cohort 1 analyzed disease-free survival in cohort 1, for which we have human breast tumors at different stages and different histo- clinical follow-up. In this cohort, there was a poor prognosis for the pathological grades and compared them with normal glands patients with tumors with low Spn with a higher number of

Figure 1. Loss of Spn in human tumors. (a) Representative images of different breast tumors with different Spn levels. (b) Percentage of human breast tumors showing high, low or no levels of Spn; N = 251 (cohort 1). (c) Statistical relationship between Spn levels and the histological grade of breast tumors; N = 251 (cohort 1). Spn intensity was valuated by three independent observers and the average of the mark given, used for the graph and calculations. (d) Representative image of Spn staining of two tumors showing different differentiation degrees.

Figure 2. Spn expression in human breast tumors. (a) Spn mRNA downregulation occurs in ER-negative tumors. N = 84 (cohort 2). (b) Triple-negative tumors contain lower levels of Spn mRNA than luminal tumors; N = 84 (cohort 2). (c) Her2+ (ER/PR negative) tumors showed the lowest levels of Spn mRNA; N = 84 (cohort 2). (d) Patients with tumors with lower levels of Spn showed worse disease-free survival than patients with high levels of Spn; N = 251 (cohort 1).

fi recurrences, especially in the rst 100 months after treatment Table 1. Inverse relationship between the percentage of Spn-positive (Figure 2d). These data correlated with the observed advanced cells and CD44+/CD24 − cells in human breast tumor cells stages of colorectal carcinoma,21 and suggests that Spn loss is related to a poor prognosis in human tumors. Because a poor Case Tumor type Spn (%) CD44 ± /CD24 − (%) prognosis has been related to a lack of therapeutic response Tumor 1 Luminal 93.8 0.011 of cancer stem cells contained in a tumor, and triple-negative o breast tumors are also associated with an increased number of Tumor 2 Luminal 73.3 0.001 Tumor 3 Luminal 59.5 0.041 cancer stem cells, we measured whether there was a functional Tumor 4 Luminal 57.6 0.079 correlation between lack of Spn and the CSC phenotype. Tumor 5 Triple negative 12.1 0.096 Tumor 6 Triple negative 4.92 1.78 Loss of Spn correlated with increased levels of CD44+/CD24 − Tumor 7 Triple negative 4.38 6.21 cells in vivo To study the possible relevance that Spn loss might have in the cancer stem phenotype of breast tumor in patients, we first correlated the levels of Spn protein with the percentage of CD44 correlation between Spn levels of expression and percentage +/CD24 − cells because these cells have been proposed to of CD44+/CD24 − cells in each sample showed a clear association be cancer-initiating cells in breast tumors.30–33 To that end, between tumors with higher levels of Spn and a lower percentage we collected several fresh human tumor samples (Table 1), of CD44+/CD24 − cells, and vice versa (Figure 3b). disaggregated them and analyzed the levels of Spn protein in the To fully confirm this relationship, we selected CD44+/CD24 − population by flow cytometry. We found that the different cells from the population by flow cytometry and pooled samples contained different levels of Spn protein ranking from the remaining cells of each sample. Then, we extracted 100% (maximum levels observed in these samples) to the mRNA and analyzed the levels of Spn mRNA expression barely detectable levels (o5%). These same samples were in each population. We found that CD44+/CD24- cells expressed analyzed by the percentage of cells being positive for CD44 and lower levels of Spn mRNA than the rest of the somatic cell negative for CD24 markers by flow cytometry (Figure 3a); the population (Figure 3c), confirming the relationship between Spn percentage of cells showing these characteristics ranked from expression and number of CD44+/CD24 − cells in the 6.21% to barely detectable (o0.001%). A comparison of the breast tumor.

Figure 3. Human breast cancer tumors with low Spn expression levels contain more CD44+/CD24 − cancer stem cells. (a) Two representative human breast cancer tumors, one with low Spn expression levels and other with high Spn expression levels (see Table 1), as shown in the Fluorescence-activated cell sorting (FACS) histograms. Dot blots show that tumor 7, which expressed low percentage of high Spn-expressing cells, contains more CD44+/CD24 − cells. (b) Seven human breast cancer tumors with different Spn expression levels were analyzed by FACS. The statistical analysis shows a correlation between the low levels of Spn and increased levels of CD44+/CD24 − cells. (c) Levels of Spn mRNA in selected CD44+/CD24 − stem cells are lower than in somatic cells. *Po0.05.

Absence of Spn enhances the cancer stem properties of breast Human breast epithelial populations can be maintained in tumor cells serum-free suspension cultures. The clusters of cells obtained have To determine the effect of Spn downregulation in breast tumor been called 'mammospheres'.34 These mammospheres display cells, we overexpressed Spn shRNA in T47D (from breast ductal some self-renewal ability upon disaggregation and are enriched carcinoma positive for a variety of hormone receptors) and for multi-potent epithelial progenitors.35 The mammospheres MDA-MB-468 (derived from metastatic triple-negative breast show an increase in CD44+/CD24 − cells, an overexpression cancer) cells and generated subpopulations carrying reduced levels of neo-angiogenic and cytoprotective factors and the expression of Spn. In both cell lines, we observed a 50% reduction in Spn of CSC markers.36 Therefore, we measured whether the down- protein and mRNA upon expression of the shRNA (Figures 4a and b). regulation of Spn alters the ability of breast cancer cell lines To analyze whether loss of Spn had any causal effect on to form mammospheres. the cancer stem cell phenotype, we measured two cellular Cells with reduced levels of Spn by shRNA were subjected to properties associated with the cancer-initiating cell phenotype: disaggregation, and the resulting single-cell suspension was the ability to grow tumorspheres and holoclones. plated at a clonal density (1000 cells/ml) in serum-free medium

Figure 4. Downregulation of Spn increases cancer stem cell properties. (a) Downregulation of Spn mRNA in breast tumor cells by expression of specific shRNA. Cell lines were infected with retroviruses carrying Spn shRNA or an empty vector (pRetroSuper, pRS). After selection, cells were grown to 80% confluence, and mRNA was extracted. The figure shows the results of RT–PCR of the Spn transcript as indicated in the Materials and Methods section. (b) Downregulation of Spn protein in breast tumor cells by expression of specific shRNA. Cell lines were infected with retroviruses carrying Spn shRNA or an empty vector (pRetroSuper, pRS). After selection, cells were grown to 80% confluence, and the protein was extracted. The figure shows the results of a western blot of the Spn transcript as indicated in the Materials and Methods section. (c, d) Loss of Spn increases mammosphere formation in tumor cells. Breast tumor cell lines were infected with retrovirus carrying Spn shRNA or an empty vector. After selection, 10 × 4 cells were grown in 24-well plates in the indicated conditions for the formation of mammospheres. After 10 days of culture, the number of mammospheres was counted. (c) Representative images of cultures. (d) Average of three independent experiments performed in triplicate. The bars show s.d. (e, f) loss of Spn increases holoclone formation in tumor cells. Breast tumor cell lines were infected with retrovirus carrying Spn shRNA or an empty vector. After selection, 10 × 4 cells were grown in 10-cm dishes in normal conditions. After 10 days of culture, the number of holoclones and paraclones were counted. (e) The percentage of the colonies according to phenotype. In each case, more than 100 colonies were counted. *Po0.05. (f) Representative images of holoclones (H) and paraclones (P). Holoclones were picked individually and grown into multiple colonies, indicating the colony-initiating ability. Paraclones were also picked but failed to grow when subcultivated independently. *Po0.05. supplemented with EGF, bFGF and insulin. In both cell lines, downregulated Spn showed an increased number of the the number of mammospheres generated was increased in the total number of growing colonies than parental cells population with reduced levels of Spn (Figures 4c and d), with (Supplementary Figure S3). The phenotypic characteristics of the a clearer effect in MDA-MB-468 cells. In T47D cells, the continuous individual colonies were analyzed and counted. We observed that subcultivation of mammospheres increased the cellularity Spn downregulation triggered a significant increase in the (Supplementary Figure S2) of the mammospheres as an indication percentage of holoclones in the cell lines tested (Figure 4e). of the increased proliferation of this population. To confirm the ability of indefinite growth of each phenotype, 20 Cell lines generated from carcinomas consistently produce colonies of each type were selected individually, seeded in vitro colony patterns similar to those produced by the stem in individual wells and grown in full media for 10 days. Colonies cells of normal epithelia. From the differing types of colony identified as holoclones showed continuous growth and filled the morphologies formed, it is possible to predict certain stem cell plate, whereas colonies identified as paraclones did not show characteristics.37 Several findings now suggest that malignant cell continuous growth (Figure 4f). lines may in fact retain stem cell patterns of behavior.38–41 For Our data show that Spn downregulation increases the stem-like example, malignant cell lines derived from holoclones contain capability of breast tumor cells. subpopulations of cells with putative stem cell characteristics.37,41 Therefore, we measured whether Spn downregulation alters Absence of Spn increases factors associated with stem properties the ability of breast cancer cell lines to form holoclones. It has been shown that the CSCs, or cancer-initiating cells, show an Cells with Spn downregulated by shRNA were subjected to increase in CD44+/CD24 − cells and stem cell transcription factors disaggregation, and the resulting single-cell suspension was such as OCT4, Nanog, Sox2 or KLF4.35,36 Therefore, we analyzed plated at a clonal density (1000 cells/ml) in a 10-cm dish. the effect of Spn downregulation on these stem cell factors. After 7–10 days, we measured the number of colonies and Cells depleted of Spn by shRNA were subjected to disaggrega- the phenotypic characteristics of these colonies. Cells with tion, and the resulting single-cell suspension was analyzed

© 2016 Macmillan Publishers Limited Oncogene (2016) 2777 – 2788 Loss of Spn increases the cancer stem cell population in breast tumors I Ferrer et al 2782 by Fluorescence-activated cell sorting to measure the percentage downregulating Spn was specific to the CD44+/CD24 − cancer of CD44+/CD24 − cells. Spn downregulation clearly increased the stem-like cell subpopulation. percentage of CD44+/CD24 − cells in both cell lines (Figure 5a), Next, we wished to analyze the tumorigenicity of these CD44 especially in MDA-MB-468 cells. +/CD24 − subpopulations because they carry different levels of Next, we analyzed in the whole population the levels of Nanog, stem transcription factors. We selected CD44+/CD24 − cells from Sox2, KLF4 and OCT4 stem cell transcription factors by quantitative MDA-MB-468 parental or MDA-MB-468 cells with Spn down- RT–PCR. Cells with reduced levels of Spn exhibited increased mRNA regulated, and 1000 cells were seeded in soft agar to measure levels of these factors (except for OCT4 in MDA-MB-468), with a clear their ability to grow in soft agar, a surrogate assay to in vivo (3–8-fold) increase in Sox2 (Figure 5b), an important cancer-specific tumorigenicity. We found that the soft agar-growing colonies regulator of tumor-initiating potential in epithelial tumors.42,43 were larger and in higher number in CD44+/CD24 − cells from We next explored the effect of Spn downregulation in different Spn-downregulated breast tumor cells (Supplementary Figure S4). subpopulations of cultured cells. We selected CD44+/CD24 − cells In summary, these results showed that, in human breast cells, − from control cells or those with reduced levels of Spn Spn downregulation increases the CD44+/CD24 CSC-like sub- and analyzed whether differences exist regarding stem cell population and induces the expression of breast stem cell markers transcription factors (Figure 5c). We found that CD44+/CD24 − and transcription factors. cells with reduced levels of Spn contained higher levels of Sox2, Klf4, Nanog and Oct4 mRNA than their counterpart CD44 Ectopic overexpression of Spn reduced growth and stem +/CD24 − cells from control/parental cells (Figure 5c). properties of breast tumor cells However, in the rest of the population (CD44+/CD24+ and Next, we decided to test whether Spn overexpression might CD44 − cells), the levels of stem cell factors were similar regardless influence the malignant behavior of tumor cells. To that end, of Spn levels. These results indicate that the effect of we constitutively overexpressed Spn cDNA into breast tumor cell

Figure 5. Downregulation of Spn increases cancer stem cell markers. Cell lines were infected with retroviruses carrying Spn shRNA or an empty vector (pRetroSuper, pRS). After selection, cells were grown to 80% confluence, and mRNA was extracted or the population measured by analytical FACS to identify the CD44+/CD24 − subpopulation. (a) Downregulation of Spn mRNA in breast tumor cells increases the percentage of CD44+/CD24 − cells. A total of 106 cells growing exponentially were labeled with primary antibodies (anti-CD44-APC; anti-CD24-PE) for 30 min at 4 ºC. After washing, cells were re-suspended in PBS–FBS–EDTA, stained with DAPI for 10 min at 4 °C and analyzed with the Gallios Flow Cytometer analytical FACS. The graph shows the average of five independent experiments. (b) Downregulation of Spn increases cancer stem cell markers. After selection, cells were grown to 80% confluence, and mRNA was extracted. The figure shows the results of q-RT–PCR of the different stem marker transcripts. The graphs show the average of five independent experiments (performed in triplicate). *Po0.05; **Po0.01. (c) Downregulation of Spn increases cancer stem cell markers in the CD44+/CD24 − subpopulation. Cell lines were infected with retroviruses carrying Spn shRNA or an empty vector (pRetroSuper, pRS). After selection, cells were grown to 80% confluence, and the CD44+/CD24 − subpopulation selected by sorting and separated from the rest of cells from the culture (CD44 − and CD44+/CD24+). mRNA was extracted from the different subpopulations and the levels of mRNAs of different stem cell factors measured by q-RT–PCR. The data show the average of five independent experiments measured in triplicate. *Po0.05; **Po0.01.

Oncogene (2016) 2777 – 2788 © 2016 Macmillan Publishers Limited Loss of Spn increases the cancer stem cell population in breast tumors I Ferrer et al 2783 lines. In all cell lines, Spn ectopic expression produced a significant Oct4 mRNA than their counterpart CD44+/CD24 − cells from reduction in colony number (Figure 6a). Furthermore, we control/parental cells. However, in the rest of the population generated a population of T47D cells overexpressing Spn (CD44+/CD24+ and CD44 − cells), the levels of stem cell factors (Figure 6b), but we were unable to generate MDA-MB-468 cells were similar regardless of the PPP1CA levels. overexpressing Spn despite antibiotic resistance in a few clones. These results indicate that the effect of downregulating PPP1CA The analysis of stem properties of these cells showed that Spn mimics the downregulation of Spn regarding its effect on the enforced expression produced a reduction in CD44+/CD24 − cells stem cell factors and the CD44+/CD24 − cancer stem-like cell and Sox2 mRNA expression (Figures 6c and d) as well as in the subpopulation. number of holoclones (Figure 6e). To explore the phenotypic alterations induced by down- These data indicate that Spn might function as a tumor regulation of PPP1CA, cells were subject to disaggregation, and suppressor by reducing the stemness properties of tumor cells, the resulting single-cell suspension was plated at a clonal density thus reducing the ability of the culture to propagate. (1000 cells/ml) in a 10-cm dish. After 7–10 days, we measured the number of colonies and the phenotypic characteristics of these Reduction of PPP1CA levels mimics the reduction of Spn levels in colonies. Cells with PPP1CA downregulated showed an increased culture number of growing colonies compared with parental cells Because the main partner of Spn is PPP1CA, the catalytic subunit (Figure 8a). Furthermore, the phenotypic characteristics of the of PP1a, we designed a set of experiments to explore the link individual colonies were analyzed and counted. We observed that between Spn and PP1a. To that end, we reduced the levels PPP1CA downregulation triggered a significant increase in the of PPP1CA, the catalytic subunit of PP1a and binding partner of percentage of holoclones in the cell lines tested (Figure 8b). Spn, with a shRNA targeting PPP1CA.44 Upon the expression and Together, our results indicate that reduction of PPP1CA levels selection of the shRNA against PPP1CA, both cell lines reduced mimics the reduction of Spn levels in culture. PPP1CA mRNA levels compared with the cells expressing only the vector (Figure 7a). As in the case of Spn shRNA, both cell lines DISCUSSION with reduced levels of PPP1CA increased the levels of stem cell transcription factors (Figure 7b), although these effects are more Spn is downregulated in ~ 15% of breast carcinomas, and its loss evident in T47D cells at the whole population level. We also correlates with more aggressive tumor phenotypes and hormone observed a clear increase in the percentage of the CD44+/CD24 − receptor loss. The artificial downregulation of Spn by shRNA CSC-like subpopulation (Figure 7c). Finally, we sorted both increased the tumorigenic properties of breast tumor cells populations and selected CD44+/CD24 − cells from control cells by increasing their stem-like properties and stem-related or cells with reduced levels of PPP1CA and analyzed whether transcription factors. In contrast, ectopic overexpression of Spn differences existed regarding stem cell transcription factors cDNA reduced the growth and stem-like properties of breast (Figure 7d). We found that CD44+/CD24 − cells with reduced tumor cells. This effect appears to be mediated by PP1a because levels of PPP1CA contained higher levels of Sox2, Klf4, Nanog and the downregulation of the partner of Spn, PPP1CA, mimics

Figure 6. Ectopic overexpression of Spn decreases stem properties and markers. The T47D breast tumor cell line was infected with retrovirus carrying Spn cDNA or an empty vector (v). After selection, the cells were processed as before. (a) Overexpression of Spn decreases the colony- formation efﬁciency in several breast tumor cells. Graph shows the average of three independent experiments. (b) Immunoblot indicating the overexpression of Spn protein was achieved in T47D cells. (c) Overexpression of Spn decreases the percentage of CD44+/CD24 − cells in T47D cells. The graph shows the average of three independent experiments. The bars show the s.d. (d) Overexpression of Spn decreases Sox2 transcript in T47D cells. The graph shows the average of ﬁve independent experiments. Bars show s.d. (e) Ectopic overexpression of Spn decreases holoclone formation in tumor cells. T47D breast tumor cell lines were infected with retrovirus carrying Spn cDNA or an empty vector. After selection, 10 × 4 cells were grown in 10-cm dishes in normal conditions. After 10 days of culture, the number of holoclones and paraclones were counted. The graph shows the percentage of the colonies according to phenotype. In each case, more than 100 colonies were counted. *Po0.05.

0.014 T47D 0.012 *

1.4 T47D 1.6 MDA-MB-468 CT Δ 0.01 1.2 1.4 1.2 1 0.008 1 * * pRS 0.8 0.006 0.8 shRNA 0.6 0.6 0.004 0.4 0.4 0.002 0.2 mRNA levels 2e PPP1CA mRNA 0.2 * 0 0 0 pRs PPP1CA pRs PPP1CA Nanog Sox2 KLF4 OCT4 shRNA shRNA

0.01 4 3.5 MDA-MB-468 * T47D CT 3 MDA-MB-468 Δ 0.008 pRS 3 2.5 0.006 shRNA 2 2 1.5 0.004 * 1 1 0.002 0.5 CD44+/CD24- (%) 0 0 mRNA levels 2e 0 pRS PPP1CA pRS PPP1CA Nanog Sox2 KLF4 OCT4 shRNA shRNA

5 4 6 KLF4 4 NANOG * SOX2 * OCT4 * 4 5 * 3 3 4 3 2 3 2

T47D 2 2 1 1 1 1

Relative mRNA levels 0 0 0 0 pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA shRNA shRNA shRNA shRNA shRNA shRNA shRNA shRNA CD44+/CD24- other CD44+/CD24- other CD44+/CD24- other CD44+/CD24- other

7 KLF4 35 7 * SOX2 ** 6 NANOG 6 1.2 30 OCT4 5 5 25 4 4 0.8 20 3 3 15 2 2 0.4 10

MDA-MB-468 1 1 5 0 0 0 0 Relative mRNA levels pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA pRs PPP1CA shRNA shRNA shRNA shRNA shRNA shRNA shRNA shRNA CD44+/CD24- other CD44+/CD24- other CD44+/CD24- other CD44+/CD24- other Figure 7. Downregulation of PPP1CA increases cancer stem cell properties. (a) Downregulation of Spn mRNA in breast tumor cells by expression of specific shRNA. Cell lines were infected with retroviruses carrying PPP1CA shRNA or an empty vector (pRetroSuper, pRS). After selection, cells were grown to 80% confluence, and mRNA was extracted. The figure shows the results of RT–PCR of the PPP1CA transcript. (b) Downregulation of PPP1CA increases cancer stem cell markers. After selection, cells were grown to 80% confluence, and mRNA was extracted. The figure shows the results of q-RT–PCR of the different stem marker transcript. The graphs show the average of three independent experiments (performed in triplicate). *Po0.05; **Po0.01. (c) Downregulation of PPP1CA mRNA in breast tumor cells increases the percentage of CD44+/CD24 − cells. A total of 106 cells growing exponentially were labeled with primary antibodies (anti-CD44-APC; anti-CD24-PE) for 30 min at 4 ºC. After washing, the cells were re-suspended in PBS–FBS–EDTA, stained with DAPI for 10 min at 4 °C and analyzed with the Gallios Flow Cytometer analytical FACS. The graph shows the average of three independent experiments. (d) Downregulation of PPP1CA increases cancer stem cell markers in the CD44+/CD24 − subpopulation. The cells were infected with retroviruses carrying PPP1CA shRNA or an empty vector (pRetroSuper, pRS). After selection, the cells were grown to 80% confluence, and the CD44+/CD24 − subpopulation selected by sorting and separated from the rest of cells from the culture (CD44 − and CD44+/CD24+). mRNA was extracted from the different subpopulations, and the levels of mRNAs of different stem cell factors were measured by q-RT–PCR. The data show the average of three independent experiments measured in triplicate. *Po0.05; **Po0.01.

the effect of Spn downregulation. The downregulation of encompass a higher percentage of precursors than the Spn or PPP1CA may thus increase the proportion of cancer- putatively less aggressive counterparts. initiating cells inside a tumor, suggesting that cells without Spn mRNA is also lost in a percentage of lung adeno- Spn are characterized by aggressiveness features and may carcinomas.18 Furthermore, lower levels of Spn mRNA correlated

Figure 8. (a) Increased number of colonies in cells with lower levels of PPP1CA when seeded at low density. (b) Reduction of PPP1CA increases holoclone formation in tumor cells. Breast tumor cell lines were infected with retrovirus carrying PPP1CA shRNA or an empty vector. After selection, 10 × 4 cells were grown in 10-cm dishes in normal conditions. After 10 days of culture, the number of holoclones and paraclones were counted. (e) Percentage of the colonies according to phenotype. In each case, more than 100 colonies were counted. *Po0.05. with a higher grade of renal carcinomas, ovarian carcinoma and Nanog and KLF4. In contrast, Spn overexpression reduced these chronic myelogenous leukemia.4 Low Spn levels were also stem cell transcription factors. Furthermore, Spn loss increases associated with high proliferative recurrences and poor patient CSC-like properties such as mammosphere or holoclone forma- prognosis in hepatocarcinoma, head and neck cancer and tion, whereas an ectopic increase of Spn reduced these growth advanced stages of colorectal carcinoma.19–21,45 In addition, and stem-like properties. Therefore, loss of the putative tumor in colorectal carcinomas, BRCA1 loss is not related to poor suppressor Spinophilin increases the CD44+/CD24 − cancer stem prognosis, confirming that the association relates only to Spn.21 cell population. The selection of the CD44+/CD24 − by sorting Recently, it has been shown that the reduction of Spn levels in two showed that the effect of Spn downregulation on stem cell-like colorectal carcinoma cell lines led to an increased formation characteristics occurs only in this subgroup of cells, but not in the of tumor spheres, an increased number of CD133-positive cells rest of the population without cancer-initiating cell characteristics. and an increased resistance to 5-flourouracil.45 All together, these results suggest an indirect mechanism that In human breast tumors, Spn is lost or reduced in ~ 15% of increases stem cell properties, along with the percentage of these samples, correlating with higher grade and more aggressive cells, in the CD44+/CD24 − cancer stem cell-like population, neoplasms; this loss correlates with a higher level of putative but does not affect this mechanism in non-CSC cells. CSC-like cells. According to our data, because only a partial increase Self-renewing activity and extensive proliferation, as well in CSC-like cells results in the whole cell culture, Spn loss may as undifferentiated status and capability to differentiate into result in a partial expansion of certain progenitor pools, such as heterogeneous mature cell types, represent hallmarks of cancer those giving origin to triple-negative tumors. Our data on the stem cells in culture. In breast cancer, the combination of CD44 CD44+/CD24 − cancer stem cell population suggest that +/CD24 − has emerged as one of the key markers for the the expanded pool is a predetermined subpopulation inside the identification and isolation of CSC. Cells with these markers are culture more than the result of stochastic alterations. present in secondary metastatic sites and showed high resistance What is the molecular mechanism? It has been reported that to chemotherapy and radiotherapy.46–48 Therefore, we evaluated Spn regulates PP1 activity,22 and loss of Spn reduces PPP1CA the expression of CD44 and CD24 by flow cytometry. In our hands, activity on its target pRb,23 thereby maintaining higher levels lower levels of Spn correlated with a higher proportion of CD44 of phosphorylated pRb and inducing an increased proliferative +/CD24 − cells in breast tumors. More interestingly, the ectopic response24 (Supplementary Figure S5). The pRb pathway controls downregulation of Spn and PPP1CA increased the proportion several aspects of stem cell biology, including the tight control of CD44+/CD24 − cells in culture. This effect was accompanied by of self-renewal characteristics of progenitor cells.26,27 However, an increase in stem cell transcription factors such as Sox2, Oct4, MDA-MB-468 cells do not have pRb protein. Therefore, an

© 2016 Macmillan Publishers Limited Oncogene (2016) 2777 – 2788 Loss of Spn increases the cancer stem cell population in breast tumors I Ferrer et al 2786 alternative mechanism is needed to explain the effect, such as the in the mammary glands, which form tumors in the presence of involvement of the other pocket proteins, p130 or p107, acting in mutated p53.25 place of pRb in its absence. In summary, we found that Spn mRNA and protein are lost in a In any case, it is remarkable that PP1 downregulation in our percentage of human breast tumors, especially ER-negative study mimicked the results of Spn downregulation, which can tumors. Lower levels of Spn correlated with a higher histological have targets other than pRb. grade. We also show that reduction of Spn correlates with In line with our results, Spn can restrain the self-renewal of − 28 an increase in CD44+/CD24 breast cancer cells, which have brain tumor-initiating cells and anchorage-independent growth fi 29 been prospectively identi ed as tumor-initiating cells, displaying of glioma cell lines. In addition to the PP1 regulatory activity of stem/progenitor cell properties. Furthermore, downregulation of Spn on pRb phosphorylation, PP1 has other targets, such as Spn by shRNA induces stem-like properties and a higher doublecortin, an actin-binding protein with an established role in 2,49 proportion of putative CSC-like cells in the population. Therefore, the subcellular targeting of PP1. Spn enhances PP1-mediated dephosphorylation of the PSer297 site of doublecortin.50 we suggest that Spn may be a novel candidate tumor suppressor Doublecortin is a microtubule-binding protein that induces at 17q21.33 because of its ability to control the cancer-initiating growth arrest at the G2–M phase of the cell cycle in glioma cells cell phenotype. and suppresses tumor xenograft growth in a Spn-dependent 29 manner, concomitant with PP1 localization into the cytosol. MATERIALS AND METHODS Doublecortin significantly reduces self-renewal of brain tumor stem cells in human primary glioma cells from surgically removed Cell lines and patient cohorts human glioma specimens and glioma cells in vitro and in vivo.28 T47D and MDA-MB-468 cell lines were obtained from the ECACC This effect on the restriction of self renewal of brain tumor- commercial repository. No further authentication was conducted by the initiating cells appears dependent on Spn expression.28 To what authors. Cells were negative for mycoplasma. extent there are two independent effects, on pRb and double- In the present work, we analyzed two different cohorts of patients, cortin, or additive or synergistic effect, must be further studied. whose main characteristics are summarized in Supplementary Table S2 The loss of the putative tumor suppressor Spn increases (cohort 1) and Supplementary Table S3 (cohort 2) described in the the cancer stem cell-like population; this occurred equally in supplementary materials. The cohorts were similar in terms of clinico-pathological features; ER-positive and -negative cell lines. Our tumor-initiating cells with however, due to sample availability, cohort 1 was used for studies with reduced levels of Spn are mainly negative for estrogen protein by immunohistochemistry, whereas cohort 2 was used for the or progesterone receptors. Interestingly, there was a correlation mRNA studies. between the CD44+/CD24 − phenotype and ER-negative profile in normal breast epithelial cells, which is in agreement with the 31–33 Cohort 1. Cohort 1 comprised 251 patients with surgically resected stage progenitor ability of these cells. I, II and III breast tumors, which were mostly (85%) invasive ductal In human breast tumors, Spn loss is associated with carcinomas. Half of the patients received chemotherapy following the ER-negative, triple-negative tumors, which exhibit a worse prog- surgery. Tissue arrays were generated with formalin-fixed, paraffin- nosis; this result may explain the lower overall survival of the embedded surgical primary tumor specimens for immunohistochemical patients with lower levels of Spn. Furthermore, analysis of a small staining. All patients had a minimum of 5 years follow-up. pool of ER − /Pr − /Her2+ tumors shows that these highly aggressive tumors have lower levels of Spn mRNA, correlating with a shorter Cohort 2. Cohort 2 was used to study Spn mRNA levels. We obtained life expectancy. It has been suggested that relapse and poor fresh-frozen samples from 84 invasive breast carcinomas: 50 estrogen response to chemotherapy is related to the number of CSCs. receptor-positive (luminal) tumors, 23 triple-negative tumors and 11 HER2 The fact that loss of Spn increases the stem-like properties invasive breast carcinomas. The clinical data of the tumor samples are of tumor cells may explain its association with more aggressive provided in Supplementary Table S2. Breast carcinomas were classified tumors and poor response in patients. into four IHC-based subtypes as follows: luminal A (ER+ and/or PR+, − o − The Spn gene is located in chromosome 17, at position HER2 , Ki-67 index 14%); luminal B (ER+ and/or PR+, HER2 , Ki-67 ⩾ − − − − − 17q21.33, in a cytogenetic area frequently associated with index 14%); HER2 (ER or ER+, PR , HER2+); and TN (ER ,PR ,HER2 ). microsatellite instability and LOH. This region is comprised All samples were obtained from the archives of the Department of Pathology, Hospital Virgen del Rocío (Sevilla, Spain), Hospital La Paz, of a relatively high density of tumor suppressor genes, including 4 (Madrid, Spain) or the CNIO tumor bank. In addition to tumor samples, we Spn, which may be an additional tumor suppressor as suggested used four non-neoplastic breast tissue samples to analyze the expression by genetic linkage studies. LOH in chromosome 17q21.33 has pattern of spinophilin in normal breast tissue. been observed in different tumors, including hormone-dependent This study was performed in accordance with the standard ethical and -independent tumors. Classical cytogenetic studies investigat- procedures dictated by Spanish law (Ley de Investigación Orgánica ing the genetic linkage of breast and ovarian cancers to Biomédica, 14 July 2007) and was approved by the ethics committee of the chromosome 17q have suggested the existence of a tumor Hospital Virgen del Rocío and the Fundación Pública Andaluza para suppressor gene distal to BRCA111,12 (Supplementary Table S3). la Gestión de la Investigación en Salud de Sevilla (FISEVI), Spain. Written In studies of hereditary breast cancer,13 the maximum LOD scores informed consent was obtained from all patients, and all clinical analyses obtained for an Spn locus were 5.44 in a study of 15 families and were conducted in accordance with the principles of the Helsinki in an extensive analysis of 271 breast and breast–ovarian cancer Declaration. families the LOD score was 21.68,15,16 indicating the importance of this locus, independent of BRCA1, in the pathology of these Tissue microarray immunohistochemistry hereditary cancers. Another study that investigated the correlation Immunohistochemical studies were performed in breast tumor specimens between p53 abnormalities and allelic loss of BRCA1, BRCA2 and included in a tissue microarray. This tissue microarray was constructed as adjacent loci in breast cancer found a strong correlation when previously described in Molina-Pinelo et al.18 and was processed for 14,51 using a specific marker for Spn locus. Spn may be involved immunostaining and development as in Guijarro et al.52 Primary antibodies in the tumorigenic process by functioning as a tumor suppressor (anti-spinophilin: AB5669 from Chemicon; anti-p53: p53 FL 393 (sc-6243) (47.1% LOH) in association with the loss of p53, but not in from Santa Cruz) were incubated for 40 min. We quantified the levels 14 association with Brca1, the neighboring gene. Confirming these of Spn by immunostaining and using a double-blind observation where data is the causal effect of Spn loss alone in vitro (this work) and we assigned discrete values between 0 (no Spn expression) and 2 in vivo, showing that Spn-null mice show benign hyperplasias (high expression, similar to normal glands). The values were averaged.

Oncogene (2016) 2777 – 2788 © 2016 Macmillan Publishers Limited Loss of Spn increases the cancer stem cell population in breast tumors I Ferrer et al 2787 Spn mRNA expression in frozen samples from human tumors HS00261636-n1; PPP1CA: Hs00267568_m1 Cat. #4331182. Experiments were Total RNA was extracted from selected areas of frozen tissue of breast repeated a minimum of five times independently, in triplicate samples. carcinomas containing 470% tumor cells using the mirVana miRNA isolation kit (Ambion, Austin, TX, USA) after having homogenized the tissue Design of shRNA against Spn for 2 min in a lysis buffer using an Ultra-turrax (T10 Basic, IKA, Germany), shRNAs against Spn were designed using the ‘Ambion siRNA target finder’ following the manufacturer's protocol. The quantity and quality of the and the ‘Qiagen siRNA design tool’ to choose the appropriate hairpin RNA isolated was measured using a NanoDrop ND-100 spectrophotometer oligonucleotides, which were then cloned in a pRetrosuper vector. (Thermo Scientific, Waltham, MA, USA). For mRNA spinophilin gene The antisense target sequence was as follows: shRNA3 for Spn, expression analysis, the total RNA (800 ng) was reverse transcribed using TGTACTTGCGCTCCAGGGC. the High Capacity RNA to cDNA kit (Applied Biosystems, Waltham, MA, USA) according to the manufacturer's instructions. The q-RT–PCR for spinophilin was performed with gene-specific fluorescent TaqMan probes Statistical analyses (TaqMan Gene Expression Assays, Applied Biosystems), using B2M as the All statistical analyses were performed using the IBM SPSS Statistics endogenous control to normalize for variations in the quantity of the input package (19.0 for Windows). The potential association between continuous cDNA. The Applied Biosystems 7900HT Fast Real Time PCR System was or categorical variables was assessed by the Pearson correlation coefficient used under the following PCR conditions: 2 min at 50 °C, 10 min at 94.5 °C, or by the Chi-squared or Fisher's exact test, respectively. The distribution 40 × 30 s melting cycles at 97 °C and annealing for 1 min at 59.7 °C. Relative ΔΔCt of quantitative variables among different study groups was assessed mRNA expression levels were calculated using the 2- formula and SDS using parametric (Student’s t-test) or non-parametric (Kruskal–Wallis 2.3 software (Applied Biosystems). or Mann–Whitney) tests, as appropriate. The disease-free interval of patients was calculated using the Kaplan–Meier method, and differences Protein isolation and western blot analysis among subgroups were assessed by the log-rank test. Experiments were performed a minimum of three independent times and always performed Cells were lysed with RIPA buffer (50 mM Tris-HCl at pH 7.6, 150 mM NaCl, in independent triplicate samples. q-PCR were repeated a minimum of five 1% NP-40, 0.5% sodium deoxycholate, 0,1% SDS, 5 mM EDTA plus proteases and phosphatases inhibitors) for 30 min at 4 ºC. Fresh proteases and independent times in triplicate. phosphatase inhibitors (Roche, Barcelona, Spain) were added to the lysis buffer. Total protein quantification was performed with Bio-Rad Protein Cell culture, retroviral vectors and gene transfer μ Assay Dye Reagent concentrate. A total of 40 g of protein was separated Cells were cultured following the experimental procedure indicated in the – on 7.5% SDS PAGE gels at 100 V and transferred to PVDF membranes for ATCC cell line data sheet. Retroviral vectors and gene transfer were 2 h at 200 mA. Membranes were hybridized with primary antibodies performed as previously described in.18,24 Colony-formation assays were as indicated (anti-Spinophilin ab5669 Millipore, Darmstadt, Germany). conducted as previously described.24 Secondary antibodies conjugated to HRP were used, and the proteins were detected using the Bio-Rad ChemiDoc XRST. Experiments were repeated a minimum of three times independently, in triplicate samples. Holo- and paraclones analysis Holo- and paraclones analysis was performed as described in Locke et al.37 fl Fluorescence-activated cell sorting analysis Brie y, for studies of colony morphology and staining patterns, actively growing subconfluent cells were dissociated in trypsin/EDTA, seeded in Tumors were surgically removed, chopped and digested in digestion 10-cm dishes at densities ranging from 50 to 100 cells/cm2, and examined media (DMEM F12+penicillin/streptomycin+collagenase+dispase+HEPES) by phase contrast microscopy as colonies developed. Colonies classified at 37 °C with agitation for 1 h. After digestion and washing with L15 culture morphologically as holoclones and paraclones were isolated with cloning medium containing 10% FBS, red blood cells were lysed with ACK lysis rings; the cells were resuspended in medium and then seeded in 24-well buffer at room temperature for 10 min. Cell pellets were re-suspended in plates. Cultures were maintained until well-developed colonies had formed L15 culture medium containing 10% FBS and stored at 4 °C overnight. The and were then analyzed. Experiments were repeated a minimum of three day after, cells were collected, washed twice with PBS, and treated with times independently, in triplicate samples. trypsin (5 min at 37 ºC) and dispase+DNase (5 min at 37 ºC). Then, viable cells were isolated using the density separation medium Lympholyte mammal by centrifugation (800 rpm for 20 min at RT). Viable cells were Mammospheres assay counted; 105 viable cells were re-suspended in 25 μl of PBS containing Cells (1 × 103) were resuspended (in triplicate) in 1 ml of complete 2% FBS and 2 mM EDTA. Cells were blocked with human blocking reagent MammoCult medium (contains the MammoCult Basal medium, Mammo- (Miltenyi Biotec, Pozuelo de alarcon, Madrid, Spain) for 10 min at 4 ºC. Cells Cult Proliferation Supplement, fresh Hydrocortisone and Heparin; were then labeled with the primary antibody anti-Spinophilin (ab5669, Stemcell technologies) and seeded in ultra-low attachment 24-well plates – – Millipore) for 30 min at 4 ºC, washed with PBS FBS EDTA and incubated (Costar #3473). Cultures were incubated in a 5% CO2 humidified incubator with the secondary alexa-488-conjugated anti-rabbit IgG and other at 37 °C for 15 days. Fresh medium was added once a week; the cultures conjugated primary antibodies (anti-CD44-APC BD559942; anti-CD24-PE were imaged, and the mammospheres were counted at days 2, 8 and 15. BD555428) for 30 min at 4 ºC. After washing, cells were re-suspended in Experiments were repeated a minimum of three times independently, PBS–FBS–EDTA, stained with DAPI for 10 min at 4 °C and analyzed with the in triplicate samples. Gallios Flow Cytometer (Beckman Coulter, Staten Island, NY, USA). Experiments were repeated a minimum of three times independently, in triplicate samples. CONFLICT OF INTEREST The authors declare no conflict of interest. RT–qPCR RNA was isolated from cells with the TRIzol reagent according to the manufacturer's instructions. One microgram of total RNA was used to ACKNOWLEDGEMENTS generate cDNA by a reverse transcriptase reaction that was performed by AC lab was supported by grants from the Spanish Ministry of Economy and sequential incubation at 25 °C for 10 min, 37 °C for 120 min and 85 °C for Competitivity, Plan Nacional de I+D+I 2008-2011, Plan Estatal de I+D+I 2013-2016, 5 min. The PCR reaction mixture (10 μl) contained 0.66 μl of the reverse ISCIII (Fis: PI12/00137, PI15/00045, RTICC: RD12/0036/0028) co-funded by FEDER from transcriptase reaction product, 5 μl of TaqMan 2 × Universal PCR Master Regional Development European Funds (European Union), Consejeria de Ciencia e Mix and 0.5 μl of the appropriate TaqMan Assay (20 × ) containing primers Innovacion (CTS-6844 and CTS-1848) and Consejeria de Salud of the Junta de and a probe for the mRNA of interest (Applied Biosystems). The mixture Andalucia (PI-0135-2010 and PI-0306-2012). This work has been also possible with the was initially incubated at 95 °C for 10 min, followed by 40 cycles at 95 °C for help of the Plan Estatal de I+D+i 2013-2016, Grant PIE13/0004 co-funded by the ISCIII 15 s and 60 °C for 60 s. We used the following probes for q-RT–PCR: and FEDER funds. We thank the donors and the CNIO tumor bank and HUVR-IBiS GAPDH: Hs03929097_g1; BMI1: Hs00995536_m1 Cat. #4331182; POU5F1 Biobank (Andalusian Public Health System Biobank and ISCIII-Red de Biobancos (OCT-04): Hs00999632_g1 Cat. #4331182; SOX2: Hs01053049_s1 Cat. PT13/0010/0056) for the human specimens used in this study. IF was funded by a #4331182; NANOG: Hs04260366_g1 Cat. #4331182. Spn: ppp1r9b- Sara Borrel Fellowship (CD12/00596).

© 2016 Macmillan Publishers Limited Oncogene (2016) 2777 – 2788 Loss of Spn increases the cancer stem cell population in breast tumors I Ferrer et al 2788 REFERENCES 27 Santamaria D, Malumbres M. Tumor suppression by Spinophilin. Cell Cycle 2011; 1 Allen PB, Ouimet CC, Greengard P. Spinophilin, a novel protein phosphatase 1 10: 2831–2832. binding protein localized to dendritic spines. Proc Natl Acad Sci USA 1997; 94: 28 Santra M, Santra S, Buller B, Santra K, Nallani A, Chopp M. Effect of doublecortin 9956–9961. on self-renewal and differentiation in brain tumor stem cells. Cancer Sci 2011; 102: 2 Satoh A, Nakanishi H, Obaishi H, Wada M, Takahashi K, Satoh K et al. Neurabin-II/ 1350–1357. spinophilin. An actin filament-binding protein with one pdz domain localized 29 Santra M, Zhang X, Santra S, Jiang F, Chopp M. Ectopic doublecortin gene at cadherin-based cell-cell adhesion sites. J Biol Chem 1998; 273: 3470–3475. expression suppresses the malignant phenotype in glioblastoma cells. 3 Sarrouilhe D, di Tommaso A, Metaye T, Ladeveze V. Spinophilin: from partners to Cancer Res 2006; 66: 11726–11735. functions. Biochimie 2006; 88: 1099–1113. 30 Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF. Prospective 4 Carnero A. Spinophilin: a new tumor suppressor at 17q21. Curr Mol Med 2012; 12: identification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 2003; 100: 528–535. 3983–3988. 5 Vivo M, Calogero RA, Sansone F, Calabro V, Parisi T, Borrelli L et al. The human 31 Mannello F. Understanding breast cancer stem cell heterogeneity: time to move tumor suppressor arf interacts with spinophilin/neurabin II, a type 1 protein- on to a new research paradigm. BMC Med 2013; 11:169. phosphatase-binding protein. J Biol Chem 2001; 276: 14161–14169. 32 Mannello F, Ligi D. Resolving breast cancer heterogeneity by searching reliable 6 Brady AE, Wang Q, Colbran RJ, Allen PB, Greengard P, Limbird LE. Spinophilin protein cancer biomarkers in the breast fluid secretome. BMC Cancer 2013; 13: stabilizes cell surface expression of alpha 2B-adrenergic receptors. J Biol Chem 344. 2003; 278: 32405–32412. 33 Ghebeh H, Sleiman GM, Manogaran PS, Al-Mazrou A, Barhoush E, Al-Mohanna FH 7 Wang Q, Zhao J, Brady AE, Feng J, Allen PB, Lefkowitz RJ et al. Spinophilin blocks et al. Profiling of normal and malignant breast tissue show CD44high/CD24low arrestin actions in vitro and in vivo at G protein-coupled receptors. Science 2004; phenotype as a predominant stem/progenitor marker when used in combination 304: 1940–1944. with Ep-CAM/CD49f markers. BMC Cancer 2013; 13: 289. 8 Kuntziger T, Landsverk HB, Collas P, Syljuasen RG. Protein phosphatase 1 34 Dontu G, Al-Hajj M, Abdallah WM, Clarke MF, Wicha MS. Stem cells in normal regulators in DNA damage signaling. Cell Cycle 2011; 10: 1356–1362. breast development and breast cancer. Cell Prolif 2003; 36:59–72. 9 Allen PB, Zachariou V, Svenningsson P, Lepore AC, Centonze D, Costa C et al. 35 Dontu G, Abdallah WM, Foley JM, Jackson KW, Clarke MF, Kawamura MJ et al. Distinct roles for spinophilin and neurabin in dopamine-mediated plasticity. In vitro propagation and transcriptional profiling of human mammary Neuroscience 2006; 140:897–911. stem/progenitor cells. Genes Dev 2003; 17: 1253–1270. 10 Hsieh-Wilson LC, Allen PB, Watanabe T, Nairn AC, Greengard P. Characterization of 36 Ponti D, Costa A, Zaffaroni N, Pratesi G, Petrangolini G, Coradini D et al. Isolation the neuronal targeting protein spinophilin and its interactions with protein and in vitro propagation of tumorigenic breast cancer cells with stem/progenitor phosphatase-1. Biochemistry 1999; 38: 4365–4373. cell properties. Cancer Res 2005; 65: 5506–5511. 11 Porter DE, Steel CM, Cohen BB, Wallace MR, Carothers A, Chetty U et al. Genetic 37 Locke M, Heywood M, Fawell S, Mackenzie IC. Retention of intrinsic stem linkage analysis applied to unaffected women from families with breast cancer cell hierarchies in carcinoma-derived cell lines. Cancer Res 2005; 65: can discriminate high- from low-risk individuals. Br J Surg 1993; 80: 1381–1385. 8944–8950. 12 Porter DE, Cohen BB, Wallace MR, Smyth E, Chetty U, Dixon JM et al. Breast cancer 38 Hirschmann-Jax C, Foster AE, Wulf GG, Nuchtern JG, Jax TW, Gobel U et al. incidence, penetrance and survival in probable carriers of BRCA1 gene mutation in A distinct "side population" of cells with high drug efflux capacity in human families linked to BRCA1 on chromosome 17q12-21. Br J Surg 1994; 81:1512–1515. tumor cells. Proc Natl Acad Sci USA 2004; 101: 14228–14233. 13 Cohen BB, Porter DE, Wallace MR, Carothers A, Steel CM. Linkage of a major breast 39 Setoguchi T, Taga T, Kondo T. Cancer stem cells persist in many cancer cell lines. cancer gene to chromosome 17q12-21: results from 15 Edinburgh families. Cell Cycle 2004; 3:414–415. Am J Hum Genet 1993; 52:723–729. 40 Resnicoff M, Medrano EE, Podhajcer OL, Bravo AI, Bover L, Mordoh J. 14 Tseng SL, Yu IC, Yue CT, Chang SF, Chang TM, Wu CW et al. Allelic loss at BRCA1, Subpopulations of MCF7 cells separated by Percoll gradient centrifugation: a BRCA2, and adjacent loci in relation to TP53 abnormality in breast cancer. model to analyze the heterogeneity of human breast cancer. Proc Natl Acad Sci Genes Chromosomes Cancer 1997; 20:377–382. USA 1987; 84:7295–7299. 15 Smith SA, Easton DF, Ford D, Peto J, Anderson K, Averill D et al. Genetic het- 41 Barrandon Y, Green H. Three clonal types of keratinocyte with different capacities erogeneity and localization of a familial breast-ovarian cancer gene on for multiplication. Proc Natl Acad Sci USA 1987; 84: 2302–2306. chromosome 17q12-q21. Am J Hum Genet 1993; 52:767–776. 42 Boumahdi S, Driessens G, Lapouge G, Rorive S, Nassar D, Le Mercier M et al. SOX2 16 Easton DF, Bishop DT, Ford D, Crockford GP. Genetic linkage analysis in familial controls tumour initiation and cancer stem-cell functions in squamous-cell breast and ovarian cancer: results from 214 families. The Breast Cancer Linkage carcinoma. Nature 2014; 511:246–250. Consortium. Am J Hum Genet 1993; 52:678–701. 43 Siegle JM, Basin A, Sastre-Perona A, Yonekubo Y, Brown J, Sennett R et al. SOX2 is 17 Abujiang P, Mori TJ, Takahashi T, Tanaka F, Kasyu I, Hitomi S et al. Loss of hetero- a cancer-specific regulator of tumour initiating potential in cutaneous squamous zygosity (LOH) at 17q and 14q in human lung cancers. Oncogene 1998; 17: 3029–3033. cell carcinoma. Nat Commun 2014; 5: 4511. 18 Molina-Pinelo S, Ferrer I, Blanco-Aparicio C, Peregrino S, Pastor MD, Alvarez-Vega J 44 Castro ME, Ferrer I, Cascon A, Guijarro MV, Lleonart M, Cajal SR et al. PPP1CA et al. Down-regulation of spinophilin in lung tumours contributes to contributes to the senescence program induced by oncogenic Ras. Carcinogenesis tumourigenesis. J Pathol 2011; 225:73–82. 2008; 29: 491–499. 19 Aigelsreiter AM, Aigelsreiter A, Wehrschuetz M, Ress AL, Koller K, Salzwimmer M 45 Ress AL, Stiegelbauer V, Schwarzenbacher D, Deutsch A, Perakis S, Ling H et al. et al. Loss of the putative tumor suppressor protein spinophilin is associated with Spinophilin expression determines cellular growth, cancer stemness and poor prognosis in head and neck cancer. Hum Pathol 2014; 45:683–690. 5-flourouracil resistance in colorectal cancer. Oncotarget 2014; 5: 8492–8502. 20 Aigelsreiter A, Ress AL, Bettermann K, Schauer S, Koller K, Eisner F et al. 46 Pinto CA, Widodo E, Waltham M, Thompson EW. Breast cancer stem cells and Low expression of the putative tumour suppressor spinophilin is associated with epithelial mesenchymal plasticity - Implications for chemoresistance. Cancer Lett higher proliferative activity and poor prognosis in patients with hepatocellular 2013; 341:56–62. carcinoma. Br J Cancer 2013; 108: 1830–1837. 47 Shipitsin M, Campbell LL, Argani P, Weremowicz S, Bloushtain-Qimron N, 21 Estevez-Garcia P, Lopez-Calderero I, Molina-Pinelo S, Munoz-Galvan S, Salinas A, Yao J et al. Molecular definition of breast tumor heterogeneity. Cancer Cell 2007; Gomez-Izquierdo L et al. Spinophilin loss correlates with poor patient prognosis in 11:259–273. advanced stages of colon carcinoma. Clin Cancer Res 2013; 19: 3925–3935. 48 Chang CC. Recent translational research: stem cells as the roots of breast cancer. 22 Ouimet CC, Katona I, Allen P, Freund TF, Greengard P. Cellular and subcellular Breast Cancer Res 2006; 8: 103. distribution of spinophilin, a PP1 regulatory protein that bundles F-actin 49 Tsukada M, Prokscha A, Oldekamp J, Eichele G. Identification of neurabin II as a in dendritic spines. J Comp Neurol 2004; 479:374–388. novel doublecortin interacting protein. Mech Dev 2003; 120: 1033–1043. 23 Berndt N. Protein dephosphorylation and the intracellular control of the 50 Bielas SL, Serneo FF, Chechlacz M, Deerinck TJ, Perkins GA, Allen PB et al. cell number. Front Biosci 1999; 4:D22–D42. Spinophilin facilitates dephosphorylation of doublecortin by PP1 to mediate 24 Ferrer I, Blanco-Aparicio C, Peregrina S, Canamero M, Fominaya J, Cecilia Y et al. microtubule bundling at the axonal wrist. Cell 2007; 129:579–591. Spinophilin acts as a tumor suppressor by regulating Rb phosphorylation. Cell 51 Querzoli P, Albonico G, di Iasio MG, Ferretti S, Rinaldi R, Cariello A et al. Cycle 2011; 10: 2751–2762. Biophenotypes and survival of BRCA1 and TP53 deleted breast cancer in 25 Ferrer I, Peregrino S, Canamero M, Cecilia Y, Blanco-Aparicio C, Carnero A. young women. Breast Cancer Res Treat 2001; 66:135–142. Spinophilin loss contributes to tumorigenesis in vivo. Cell Cycle 2011; 10:1948–1955. 52 Guijarro MV, Leal JF, Fominaya J, Blanco-Aparicio C, Alonso S, Lleonart M et al. 26 Galderisi U, Cipollaro M, Giordano A. The retinoblastoma gene is involved in MAP17 overexpression is a common characteristic of carcinomas. Carcinogenesis multiple aspects of stem cell biology. Oncogene 2006; 25: 5250–5256. 2007; 28: 1646–1652.

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