CIP2A Signature Reveals the MYC Dependency of CIP2A-Regulated Phenotypes and Its Clinical Association with Breast Cancer Subtypes

Oncogene (2012) 31, 4266–4278 & 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc ORIGINAL ARTICLE CIP2A signature reveals the MYC dependency of CIP2A-regulated phenotypes and its clinical association with breast cancer subtypes

M Niemela¨1,2,3, O Kauko1,2,4, H Sihto5, J-P Mpindi6, D Nicorici6, P Pernila¨5, O-P Kallioniemi6, H Joensuu5, S Hautaniemi7 and J Westermarck1,4,6

1Turku Centre for Biotechnology, University of Turku and A˚bo Akademi University, Turku, Finland; 2Turku Graduate School of Biomedical Sciences, Turku, Finland; 3Department of Medical Biochemistry and Genetics, University of Turku, Turku, Finland; 4Department of Pathology, University of Turku, Turku, Finland; 5Laboratory of Molecular Oncology, Biomedicum, University of Helsinki, Helsinki, Finland; 6Institute for Molecular Medicine Finland, University of Helsinki, Helsinki, Finland and 7Research Programs Unit, Genome-Scale Biology and Institute of Biomedicine, Biochemistry and Developmental Biology, University of Helsinki, Helsinki, Finland

Protein phosphatase 2A (PP2A) is a critical human a plausible novel explanation for the high MYC activity in tumor-suppressor complex. A recently characterized basal-like and HER2 þ breast cancers. PP2A inhibitor protein, namely cancerous inhibitor of Oncogene (2012) 31, 4266–4278; doi:10.1038/onc.2011.599; PP2A (CIP2A), has been found to be overexpressed at a published online 16 January 2012 high frequency in most of the human cancer types. However, our understanding of gene expression programs Keywords: KIAA1524; breast cancer; gene expression regulated by CIP2A is almost absent. Moreover, clinical profile; PLAUR; SERPINE2; PP2A B-subunit relevance of the CIP2A-regulated transcriptome has not been addressed thus far. Here, we report a high-confidence transcriptional signature regulated by CIP2A. Bioinfor- matic pathway analysis of the CIP2A signature revealed Introduction that CIP2A regulates several MYC-dependent and MYC- independent gene programs. With regard to MYC- Previous work by Weinberg and colleagues has revealed independent signaling, JNK2 expression and transwell that regardless of the vast number of different signaling migration were inhibited by CIP2A depletion, whereas proteins that have been linked to maintenance of MYC depletion did not affect either of these phenotypes. tumorigenic phenotype, perturbation of only limited Instead, depletion of either CIP2A or MYC inhibited number of genetic elements is sufficient to induce cancer cell colony growth with statistically indistinguish- cellular transformation in many different human cell able efficiency. Moreover, CIP2A depletion was shown to types (Hahn et al., 1999; Rangarajan et al., 2004; Zhao regulate the expression of several established MYC target et al., 2004). Therefore, these common genetic elements genes, out of which most were MYC-repressed genes. could be considered as master regulators of cancer CIP2A small-interfering RNA-elicited inhibition of col- development and progression. One of the important ony growth or activation of MYC-repressed genes was conclusions of their study was that in addition to reversed at large by concomitant PP2A inhibition. inhibition of two well-established tumor suppressors, Finally, the CIP2A signature was shown to cluster with namely p53 and Rb, inhibition of protein phosphatase basal-type and human epidermal growth factor receptor 2A (PP2A) activity was also required for malignant transformation of normal human cells expressing (HER)2-positive (HER2 þ ) breast cancer signatures. Accordingly, CIP2A protein expression was significantly activated HA-Ras and telomerase reverse transcriptase (Hahn et al., 1999; Rangarajan et al., 2004; Zhao et al., associated with basal-like (P ¼ 0.0014) and HER2 þ 2004). Together with results of several follow-up studies (Po0.0001) breast cancers. CIP2A expression also (Chen et al., 2004; Zhao et al., 2004; Westermarck and associated with MYC gene amplification (Po0.001). Taken together, identification of CIP2A-driven transcrip- Hahn, 2008), these data have now established that PP2A tional signature, and especially novel MYC-independent is one of the critical human tumor suppressors (Mumby, signaling programs regulated by CIP2A, provides 2007; Westermarck and Hahn, 2008; Eichhorn et al., important resource for understanding CIP2A’s role as a 2009). PP2A is a trimeric phosphatase complex that clinically relevant human oncoprotein. With regard to accounts for a significant fraction of all cellular serine/ MYC, these results both validate CIP2A’s role in threonine phosphatase activity. The PP2A complex regulating MYC-mediated gene expression and provide consists of a catalytic C subunit (PP2Ac), a scaffolding A subunit (PR65) and various regulatory B subunits (Janssens et al., 2008; Eichhorn et al., 2009). PP2A Correspondence: Prof J Westermarck, Centre for Biotechnology, activity is regulated by its complex composition and by University of Turku, Tykisto¨katu 6A, 20520 Turku, Finland. E-mail: jukwes@utu.fi post-translational modifications of its subunits (Janssens Received 10 June 2011; revised 29 October 2011; accepted 21 November et al., 2008; Eichhorn et al., 2009). A third level of 2011; published online 16 January 2012 regulation of PP2A activity is interaction of PP2A Identification of the CIP2A-regulated transcriptome M Niemela¨ et al 4267 dimers and trimers with endogenously expressed PP2A- phenotypes regulated by CIP2A has not yet been interacting proteins (Westermarck and Hahn, 2008; thoroughly addressed. As CIP2A-regulated transcrip- Virshup and Shenolikar, 2009). Nevertheless, very little tional programs have not been characterized thus far, it is known about the mechanisms regulating PP2A has not been possible to address whether the functional complex composition and/or its activity in mammalian effects of CIP2A on MYC are linked to certain types of cells, or the clinical relevance of PP2A regulation in cancer. human malignancies. In this study, we report the first global transcriptome With regard to the mechanisms by which PP2A analysis of the CIP2A function. Bioinformatic analysis activity may be inhibited in human malignancies, we of its transcriptional signature suggested that CIP2A recently identified a novel protein–protein interaction promotes both MYC-dependent and MYC-independent partner for PP2A (Junttila et al., 2007). On the basis of signaling. Accordingly, we demonstrate a previously the PP2A-inhibiting activity of this protein and its unknown MYC-independent role for CIP2A in regulat- cancer-selective expression, it was designated as the ing JNK2 gene expression and transwell cell migration. cancerous inhibitor of PP2A (CIP2A) (Junttila et al., With regard to MYC-dependent CIP2A roles, we show 2007). CIP2A depletion results in inhibition of malig- that CIP2A regulates the expression of previously nant cell growth and inhibition of tumor growth in identified direct MYC target genes. However, the mouse xenograft models (Junttila et al., 2007; Come CIP2A signature also reveals several previously uni- et al., 2009). CIP2A is overexpressed in most major solid dentified MYC-regulated genes. Importantly, we verify human cancer types, including head and neck squamous that CIP2A’s effects on both MYC expression and cell carcinomas as well as gastric, esophageal, breast, colony growth are reversed by concomitant PP2A colon, lung and prostate cancer (Junttila et al., 2007; Li inhibition. Finally, we demonstrate that the CIP2A et al., 2008; Come et al., 2009; Khanna et al., 2009; signature clusters together with basal-like and HER2 þ Dong et al., 2010; Katz et al., 2010; Qu et al., 2010; breast cancer signatures and that the CIP2A protein Vaarala et al., 2010). Strikingly, with the exception of expression is significantly associated with both of these breast cancer, in which 40% of patient samples over- breast cancer subtypes. CIP2A is also shown to promote express CIP2A (Come et al., 2009), all other studied MYC target gene expression in basal-like breast cancer human cancer types have CIP2A overexpression in cells. Thus, our results indicate that the CIP2A signature 65–90% of patient samples. Moreover, CIP2A expression could be useful in identification of both novel surrogate predicts poor patient prognosis in certain gastric cancer markers and therapy targets in MYC-driven basal-like subtypes and in non-small cell lung cancer (Khanna et al., and HER2 þ breast cancers. 2009; Dong et al., 2010). In breast cancer, CIP2A expression correlates with disease aggressiveness and in prostate cancer with a high Gleason’s score (Come et al., Results 2009; Vaarala et al., 2010). Taken together, these studies have firmly established CIP2A as a novel, clinically Characterization of the CIP2A-driven transcriptional relevant human oncoprotein. signature The transcription factor MYC is one of the most To characterize the transcriptional signature driven by studied and important human oncoproteins (Meyer and CIP2A in cancer cells, HeLa cells were transfected with Penn, 2008). In breast cancer, MYC activity is closely either scrambled or specificity-validated CIP2A small- linked to the basal-like and the human epidermal interfering RNA (siRNA) (Junttila et al., 2007), and growth factor receptor (HER)2-positive (HER2 þ ) genome-wide gene expression profiles were analyzed 3 subtypes (Xu et al., 2010). The MYC protein is and 5 days after transfection using the Illumina overexpressed in B45% of breast tumors, whereas platform. We have previously used HeLa cells in diverse MYC gene amplification is observed only in B15% of experiments to study the function of CIP2A, and the tumors and mRNA overexpression in 22–35% of breast effects of CIP2A in HeLa cell growth both in vitro and cancers (Xu et al., 2010). These results suggest involve- in vivo, as well as in MYC expression are well ment of post-translation regulation in MYC protein established. Thus, HeLa cells were also chosen as the overexpression in breast cancer. However, the mechan- model cell line to identify CIP2A-regulated transcrip- isms that promote MYC protein expression and activity tional signature. Downregulation of CIP2A protein in breast cancer are poorly understood. In cultured cells, levels by siRNA transfection was confirmed by western inhibition of PP2A, either by viral small-t antigen or blotting at both 3 and 5 day time points (Figure 1a). The CIP2A expression, results in MYC protein stabilization microarray raw data were normalized to quartiles, and (Junttila et al., 2007). Accordingly, several recent studies the fold change for each CIP2A/scrambled sample pair have confirmed that depletion of CIP2A in was calculated. Except certain genes, depletion of human cancer cell lines results in MYC protein CIP2A did not induce drastic changes in gene expression degradation (Junttila et al., 2007; Come et al., 2009; profiles. Instead, subtle expression level changes for the Dong et al., 2010). However, it has not yet been same genes were observed in repeat experiments. Thus, determined whether CIP2A-mediated PP2A regulation to create a high-confidence CIP2A transcriptional indeed mediates CIP2A’s effects on MYC expression signature, the experiment was repeated five times and and cancer cell proliferation. Moreover, the MYC only genes that surpassed the threshold of 1.3 in all five dependency of gene expression profiles and cellular replicate sample pairs were included in the final

Oncogene Identiﬁcation of the CIP2A-regulated transcriptome M Niemela¨ et al 4268 Log2 FC: Log2 FC:

Day 3 D a y 5 Sample pair: 12345Sample pair: 12345 PRG4 1.52 PRG4 1.21 CRYAB 1.47 SERPINE2 1.15 SAT1 1.38 COL8A1 1.15 Day 3 NA 1.19 KIAA1524 -1.08 NPTX1 1.16 DCN 1.02 PDLIM7 1.12 CRYAB 0.94 KIAA1524 -1.11 FAP 0.92 LXN 1.09 IER3 0.89 Day 3 Day5 Day 5 HIST1H2BK 1.05 PDK4 -0.88 SLC12A8 1.04 SCIN -0.85 CIP2A PSG4 -1.04 S100A16 0.83 PDGFRB 1.04 SAT1 0.81 RAB31 -1.02 PLAU 0.79 Actin 112 13 9 GRSF1 -1.00 PDLIM7 0.79 MFAP5 -0.99 LUM 0.77 siRNA: LAMA1 -0.99 NA -0.73 EBI3 0.98 SERPINE1 0.69 DSC2 0.97 SCR SCR GPNMB -0.63 CIP2A CIP2A SLC22A4 0.97 CRLF1 0.61 PLAUR 0.96 SLC25A4 -0.56 ACTG2 -0.95 ACPL2 -0.52 ENO3 0.95 PDE2A -0.52 CALHM3 0.94 C4BPA 0.93 KIAA1324 0.92 SERPINE2 0.92 RHOD 0.91 5 NA -0.90 LOC342897 0.90 4 C9orf169 0.89 3 SCIN -0.87 COL8A1 0.87 2 RHOC 0.87 1 GPR183 0.86 0 PTPRB 0.86 LASP1 -0.86 -1 ACPL2 -0.84 -2 GPR37 0.82 COL15A1 0.82 -3 (Log2 Fold change) SLC22A18 0.82 -4 FAP 0.81 GRRP1 0.81 -5 NA 0.81 NA -0.80 RT-PCR validation of signature genes FAP CORO6 0.79 SAT1 CAV1 LRP4 0.79 PLAUR ACTG2 RAB31 LAMA1 MFAP5 COL8A1 CFB 0.78 SLC22A4 RARRES1 SERPINE2 SNCA 0.78 ZDHHC11 -0.78 TUBA4A 0.77 Figure 1 Identiﬁcation of CIP2A-regulated transcriptional signature. (a) Western blot analysis of CIP2A protein expression in scrambled (SCR) and CIP2A siRNA-transfected HeLa cells used for microarray analysis. (b) Venn diagram representing the overlap of the total 134 signature genes at days 3 and 5. (c) Heatmap illustrating 50 top differentially regulated genes at day 3 and all the 22 differentially expressed genes at day 5. Green indicates downregulation and red upregulation in conditions under which CIP2A is downregulated. Decreased color intensity indicates stronger differential regulation. Heatmap was built using Ashley lab Heatmap Builder. Mean log2 changes of individual signature genes in ﬁve replicate sample pairs are shown on the right side of heatmap columns. Not annotated genes are marked with ‘NA’. (d) qRT–PCR analysis of expression of selected CIP2A signature genes in HeLa cells 3 days after CIP2A siRNA transfection. Mean±s.e.m. change as compared with scrambled siRNA-transfected cells in three independent experiments is shown.

signature. On the basis of these strict criteria, the ture was further validated in MCF7 breast cancer, CIP2A-driven transcriptional signature consists of 134 and AGS gastric cancer cell line and all 9/12 and 8/12 different genes (Supplementary Table 1). Of these, 112 signature genes that were expressed in these cells genes were seen only at day 3, and 9 genes were seen at detectable level, respectively, were similarly regulated only at day 5, whereas 13 of the total 134 signature genes by CIP2A siRNA as in HeLa cells (Supplementary were regulated at both time points (Figure 1b). The Figures 1B and C). Thus, these results verify that the heatmap shown in Figure 1c depicts expression levels of CIP2A transcriptional signature presented here consists the 50 top regulated genes at day 3 and all differentially of genes that are CIP2A regulated with high confidence regulated genes at day 5. The remaining signature genes in human cancer cells. and their regulation by CIP2A are shown in Supple- mentary Table 1. Network analysis of the CIP2A signature genes To validate CIP2A-dependent regulation of signature To characterize the cellular signaling pathways in which genes, we examined the mRNA expression levels of 12 CIP2A signature genes are involved, we conducted randomly chosen signature genes in CIP2A-depleted ingenuity pathway analysis (Sigma-Aldrich). On the HeLa cells using reverse transcriptase (RT)–qPCR basis of the day-3 signature, ingenuity pathway analysis (Figure 1d). For this purpose, we used two previously revealed a total of 7 networks that were altered published siRNA sequences against CIP2A (Junttila significantly as indicated by ingenuity score 415 et al., 2007). The specificity of one of these siRNAs has (Table 1). The most significantly altered pathway at been validated by an add-back rescue experiment day 3 consisted of several genes that have been (Junttila et al., 2007). In HeLa cells, all of the previously linked to cellular migration (Table 1, quantitative RT–PCR-tested genes were regulated simi- Figure 2a). At day 3, MYC was found in network larly as on the microarray by both of the tested siRNAs number 7, as ranked by the ingenuity score (Table 1). (Figure 1d and Supplementary Figure 1A). The signa- In turn, the analysis of the day-5 signature yielded only two

Oncogene Identiﬁcation of the CIP2A-regulated transcriptome M Niemela¨ et al 4269 Table 1 Ingenuity networks ID Score Signature Molecules in the ingenuity network genes

Day 3 1 41 21 B3GNT1, C1S, CAV2, Caveolin, CD82, CFB, Chymotrypsin, Collagen Alpha1, Collagen type I, Collagen type IV, Collagen(s), CRYAB, CYTH1, DCN, DDR1, FAP, Focal adhesion kinase, Integrin, Integrin alpha 3 beta 1, LAMA1, Laminin, Mmp, MTSS1, NFkB, Pdgf, PLAU, PLAUR, PROCR, PSAT1, SAT1, SCNN1A, SERPINE2, SPINK5, TFAP2A, Trypsin 2 34 18 Alp, ALPI, Ap1, BHLHE40, CAV1, COL8A1, Creb, EBI3, ERK1/2, GDA, GPR37, GPR183, GSN, Hsp70, IER3, IGFBP4, IL1, IL12, LASP1, LDL, LRP, LRP4, MT1E, P38 MAPK, p85 (pik3r), PDGF BB, PDGFRB, Pi3-kinase, PI3 K, Pld, Ras, SCIN, SNCA, Tgf beta, TGM2 3 27 15 26 s Proteasome, ACTG2, Actin, Akt, Alpha tubulin, ATP1B1, BMF, Ck2, COL15A1, EFEMP1, ENO3, ERK, F Actin, FSH, FURIN, hCG, Histone h3, IgG, Jnk, Lh, Mapk, MAPK9, P4HA2, peptidase, Pkc(s), PP2A, PPP2R2B, RAB31, TUBA3E, TUBA4A, TUBB2A, Tubulin, TWF2, Vegf, ZDHHC11 4 23 13 ADAMTS3, ANXA11, ARHGAP29, C4BPA, Ca2 þ , CALU, CNN3, DFNA5, DMBT1, DUSP16, FPR2, GYPC, HEXA, HSPB2, hydrogen peroxide, JAG2, lipoxin A4, LXN, MB, MMP9, MPP1, NA, NEU1, NEU3, PAPPA, PCSK6, PDLIM7, RHOC, RHOD, RND1, SERPINE2, SLC22A4, TGFB1, Timp, TNF 5 22 13 ANO2, BSCL2, BTRC, C7ORF58, CDH5, CLDN12, CNTN1, CNTNAP1, CORO6, CTNNB1, DNAJA2, DSC2, F11R, FHIT, GINS3, HIRA, HIRIP3, HIST1H2BK, HIST4H4, HNF4A, HSPB8, MIR200A, MIR24-1, MIRN151, NANOG, NPTX1, NRCAM, PKP2, PTPRB, RARG, SERPINA5, SETDB1, SLC22A18, TNC, TOE1 6 18 11 COL7A1, CTSA, DLL1, EPHB6, FBLN1, FN1, FNG, GRB2, HIRA, IFNG, JAG2, LFNG, LMO7, MFAP5, MMP26, NOTCH1, OLFM1, PAX3, PRDX2, PRG4, PRSS1, PTPN22, RARRES1, retinoic acid, SERPINA5, SHB, SHKBP1, SRGN, STX16, THRAP3, TIMP4, TMOD1, TPM4, TUBB2C, VPS45 7 17 11 ACPL2, AHSG, BNIP2, CD40LG, CDC14B, CDKN2AIP, CHP, CRIP2, F11R, FBXO2, FUBP1, GBP2, GRSF1, HLA- DQA1, IFI35, KIAA1324, KIAA1524, LY6A, MIR124, MIR128-1, MX1, MYC, OXSR1, P4HA1, PPP2R4, RASSF5, RPL23, SMN1, TAP1, TMEM2, TMSB4X, TNFSF13, TP53, WISP1, WNT3A

Day 5 1 21 9 ACPL2, AMD1, beta-estradiol, CARD6, CASP1, Cbp, CDKN1A, CRLF1, CTSB, DLAT, FGF1, FGF2, GAPDH, GBP2, GEM, GPNMB, GRB2, HABP2, HLA-DQA1, IFNG, KIAA1524, KLF6, KLF10, LUM, MYC, NA, NAB2, NLRC4, PAK7, PDK4, PDLIM7, PRG4, SERPINA5, TRA2B, WISP1 2 21 9 13(S)-hydroxyoctadecadienoic acid, Akt, C1S, Collagen type I, Collagen type IV, CRYAB, DCN, F11, FAM3B, FAP, IER3, IL1, LGALS7, Mmp, MMP7, MMRN1, MRC2, NFkB, Pdgf Ab, PDGF BB, Pkc(s), Plasminogen Activator, PLAU, PPIF, PPP2R5C, SAT1, SERPINE1, SERPINE2, SLC25A4, Tgf beta, THBD, TMPRSS6, Trypsin, WISP1

Figure 2 Ingenuity pathway analysis of CIP2A signature. (a) Most significantly altered ingenuity network at day 3 time point. (b) Most significantly altered ingenuity network at day 5 time point. Green color indicates downregulated and red color upregulated signature genes. Molecules indicated with empty icons (including MYC) were not included in the input signature but are associated with the network based on ingenuity pathway analysis. ingenuity networks that were significantly associated (Figure 2b, Table 1). As expected, due to the post- with the signature. Here, MYC was found in the most translational mechanism by which CIP2A regulates significantly altered network, that is, network number 1 MYC (Junttila et al., 2007), CIP2A siRNA inhibited

Oncogene Identification of the CIP2A-regulated transcriptome M Niemela¨ et al 4270 MYC protein expression but not MYC mRNA levels in fection did not inhibit transwell migration (Figure 3a). the samples used for microarray analysis (Supplemen- The MYC-independent effects on the transwell migra- tary Figures 2A and B). A graphical illustration of all tion of HeLa cells are consistent with previously networks significantly linked to the signature (Table 1) is published observations by Cappellen et al. (2007). shown in Supplementary Figures 3 and 4. Importantly, even though CIP2A seemed to regulate To further substantiate pathway analyses, we also cell migration by MYC-independent mechanisms, the analyzed CIP2A signature genes with the gene set migration defect in CIP2A siRNA-transfected cells was enrichment analysis (GSEA) database (http://www. rescued by subsequent treatment of cells with PP2A broadinstitute.org/gsea/). Interestingly, genes upregu- inhibitor okadaic acid (Figure 3b). Therefore, these lated by CIP2A depletion were associated with ‘genes results identify transwell migration of HeLa cells as downregulated by MYC, according to the MYC Target MYC independent but PP2A-dependent cellular pheno- Gene Database’, both in the 3- and the 5-day signatures. type regulated by CIP2A. The P-values for the association were 5.58eÀ9 and To further validate the MYC-independent functions 7.77eÀ7, respectively. Moreover, related to the cellular of CIP2A, we focused on regulation of the MAPK9 migration pathway linked to CIP2A signature by (JNK2) gene, which was linked to MYC-independent ingenuity analysis, GSEA also revealed association of cellular movement functions based on the ingenuity focal adhesion phenotype with the CIP2A signature analysis (Supplementary Table 2). JNK2 has also (P-value ¼ 4.8eÀ5). previously been implicated as the cancer-promoting isoform of the JNK family of mitogen-activated protein kinases (Mathiasen et al., 2011). CIP2A depletion- CIP2A promotes MYC-independent transwell migration and JNK2 expression in HeLa cells elicited downregulation of JNK2 mRNA and protein levels was verified by RT–qPCR and western blot To explore the functional consequences of CIP2A analysis, respectively (Figures 3c and d). However, we depletion with the help of the CIP2A signature, we next found that JNK2 mRNA expression was not inhibited analyzed the molecular and cellular functions of the by MYC siRNA, even though MYC depletion inhibited signature genes. On the basis of the day-3 signature, the expression of the validated MYC target genes, E2F2 CIP2A depletion had the most significant effect on and NCL (nucleolin) (Figure 3e). cellular movement with 28 genes affected (P ¼ 1.22E-06 On the basis of these results, we identified novel to 1.33E-02) (Table 2). Alternatively, at day 5, the most MYC-independent functions for CIP2A in promoting altered function was cellular growth and proliferation, transwell migration and JNK2 expression in HeLa cells. with 12 genes affected (P ¼ 5.61E-08 to 1.48E-02) (Table 2). A strong association of the day-3 signature with genes that have established functions in cellular CIP2A, MYC, PP2A and proliferation migration and movement (Figure 2a, Tables 1 and 2) led On the basis of the network analysis, cellular growth us to further study the potential role for CIP2A in these and proliferation pathways related to CIP2A signature phenotypes. We did not observe significant effects of genes were associated with MYC (Tables 1 and 2). To CIP2A depletion on scratch wound healing or adhesion clarify the MYC dependency of CIP2A-mediated pro- of HeLa cells (Supplementary Figures 5A and B). liferation regulation, we performed a colony-formation However, CIP2A depletion did significantly inhibit assay with either CIP2A or MYC siRNA-transfected transwell migration of HeLa cells in a Boyden chamber HeLa cells. As shown in Figures 4a and b, CIP2A and assay (Figure 3a, upper panel, and Supplementary MYC siRNAs inhibited colony growth of HeLa cells Figure 5C). As already suggested by the ingenuity with statistically indistinguishable efficiency. These analysis, the migration-promoting function of CIP2A results suggest that MYC contributes to the CIP2A- does not seem to be linked to CIP2A’s role in regulating mediated stimulation of HeLa cell colony growth. MYC because MYC inhibition through siRNA trans- Recently, several independent studies have demon-

Table 2 Ingenuity molecular and cellular functions Name P-value No. of molecules

Day 3 Cellular movement 1.22E-06 to 1.33E-02 28 Cell-to-cell signaling and interaction 5.87E-06 to 1.42E-02 24 Cellular development 3.39E-05 to 1.54E-02 29 Cellular growth and proliferation 3.39E-05 to 1.30E-02 37 Cell morphology 4.21E-05 to 1.30E-02 13

Day 5 Cellular growth and proliferation 5.61E-08 to 1.48E-02 12 Cell morphology 1.07E-05 to 1.38E-02 6 Cell-to-cell signaling and interaction 1.07E-05 to 1.17E-02 5 Carbohydrate metabolism 1.39E-05 to 6.38E-03 5 Cell death 1.48E-05 to 1.38E-02 9

Oncogene Identiﬁcation of the CIP2A-regulated transcriptome M Niemela¨ et al 4271 1.6 NS 1.4 ** 1.4 1.2 1.2 1 1 0.8 0.8 0.6 0.6

(Relative values) 0.4 0.4 Transwell migration (Relative values)

0.2 Transwell migration 0.2 0 0 siRNA: CIP2A SCR MYC SCR CIP2ASCR CIP2A MYC OA B-Actin

1.2 * 0.5 JNK2 E2F2 NCL 1 JNK2 0 0.8 CIP2A 0.6 - 0.5 B-Actin MYC 0.4 -1 siRNA siRNA: CIP2A SCR JNK2 RT-PCR

(Relative values) 0.2 -1.5 0 Log2 Fold change siRNA: CIP2A SCR -2 Figure 3 CIP2A promotes transwell migration and JNK2 expression in an MYC-independent manner. (a) Upper panel: Boyden chamber transwell migration assay in HeLa cells treated with CIP2A, scrambled (SCR) or MYC siRNA. Mean±s.e.m. of four independent experiments (**Po0.01, NS, P40.05; two-tailed Student’s t-test) is shown. Lower panel: immunoblot of the MYC protein levels in cells used for the transwell migration assay. (b) Inhibition of protein phosphatase 2A activity by okadaic acid (OA) rescues inhibition of transwell migration of HeLa cells by CIP2A siRNA. Mean±s.e.m. of independent experiments is shown. (c) qRT–PCR quantification of JNK2 mRNA levels in CIP2A siRNA-treated cells. Mean±s.e.m. of three independent experiments (*Po0.05; two-tailed Student’s t-test) is shown. (d) Immunoblot of the JNK2 protein levels in CIP2A siRNA-treated cells. (e) qRT– PCR analysis of JNK2, E2F2 and NCL mRNA levels in MYC siRNA-treated cells. Mean±s.e.m. of three independent experiments is shown. strated that CIP2A inhibits cellular PP2A activity regulation of MYC protein expression is consistent (Junttila et al., 2007; Chen et al., 2010; Guenebeaud with recently published findings (Arnold and Sears, et al., 2010). Moreover, in HeLa cells used for this 2006; Ben-Israel et al., 2008; Sablina et al., 2010). study, siRNA-mediated CIP2A inhibition induced cellular PP2A activity (data not shown). To define Identification of MYC-regulated genes from the CIP2A the role of PP2A in the CIP2A-mediated regulation of signature MYC expression, we tested whether the simultaneous Results above link CIP2A and MYC together in the siRNA-based depletion of certain PP2A B subunits regulation of cancer cell proliferation. Next, the CIP2A could rescue the downregulation of MYC protein signature was used to assess the MYC dependency of levels in CIP2A-depleted cells. Out of altogether five B CIP2A’s effects on regulation of gene expression. subunits screened, we could identify two proteins To this end, we studied how the 12 RT–PCR-validated (B55a and B56b) the depletion of which invariably CIP2A signature genes shown in Figure 1d were and significantly reversed inhibition of MYC protein regulated by MYC siRNA. Among the 12 studied genes expression (Figure 4c). Efficient depletion of these B (Figure 5a), 4 are known MYC targets (PLAUR, subunits was controlled by RT–PCR (Supplementary SERPINE2, SLC22A4 and CAV1) (http://www.myc- Figure 6). Importantly, depletion of these same B cancer-gene.org). Moreover, out of these established subunits (B55a and B56b) also almost completely MYC targets, SLC22A4 and SERPINE2 have been reversed CIP2A siRNA-elicited inhibition of HeLa demonstrated to directly bind MYC at their promoters cell colony growth (Figure 4c). (Fernandez et al., 2003; Li et al., 2003). As shown in Taken together, these findings validate the signa- Figure 5a, the expression of PLAUR, SERPINE2 and ture-based prediction that the proliferation effects of SLC22A4 was upregulated after MYC depletion (M), CIP2A-mediated PP2A inhibition are mediated by and as was already shown in Figure 1d, CIP2A MYC. Moreover, these results identify PP2A com- inhibition (C) also stimulated the expression of these plexes containing the B subunits B55a,orB56b being genes. Furthermore, by analyzing a panel of human involved in MYC and proliferation regulation by breast-derived cell lines for their expression levels of the CIP2A. Identification of the involvement of several validated MYC targets, PLAUR, SERPINE2 and PP2A B subunits, including B55a, in the negative SLC22A4, we identified 15 cell lines in which inverse

Oncogene Identiﬁcation of the CIP2A-regulated transcriptome M Niemela¨ et al 4272 1.2 ** ** 1 NS 0.8

0.6

0.4 Foci formation (% of SCR siRNA) siRNA: SCR MYC CIP2A 0.2

0 siRNA: SCR MYC CIP2A

NS 1.6 NS 1.4 ** 1.2 1 0.8 0.6 0.4 (relative levels) 0.2 MYC protein expression 0 β β α siRNA: α SCR B56 B55 CIP2A CIP2A +B56 CIP2A +B55 Colony growth: 1.00 0.50 1.09 0.93 0.75 0.81 S.E.M: 0.00 0.08 0.18 0.13 0.06 0.05 Figure 4 Functional relationship between CIP2A, MYC and PP2A in regulation of cancer cell colony growth. (a) Inhibition of HeLa cell colony formation by CIP2A and MYC siRNAs. Images of the representative plates 10 days after transfection ate shown. (b) Quantification of the number of colonies in response to either MYC or CIP2A depletion. Mean±s.e.m. of three independent experiments (**Po0.01, NS, P40.05; two-tailed Student’s t-test) is shown. (c) Upper panel: quantification of MYC protein levels in HeLa cells incubated for 3 days with indicated siRNA treatments. Mean±s.e.m. of MYC/actin levels from 2 to 3 independent experiments (**Po0.01, NS, P40.05; two-tailed Student’s t-test) is shown. Lower panel: Quantification of the HeLa cell colony growth in response to indicated siRNA treatment. Mean±s.e.m. of relative number of colonies from 2 to 3 independent experiments.

correlation between the expression of CIP2A promoter (Figure 5d), demonstrating that CIP2A also (KIAA1524) and at least 2 out of 3 of these genes was promotes the function of activating MYC complexes. As observed (Figure 5b). These results indicate that CIP2A- a control, CIP2A depletion did not affect p53-regulated mediated regulation of these MYC target genes is not transcriptional activity (Figure 5d). restricted to conditions in which CIP2A expression has As pointed out above, out of a total of 134 differently been modulated by siRNA. regulated genes by CIP2A, 91 were upregulated and 43 Among the CIP2A signature genes that were not were downregulated in response to CIP2A depletion previously known to be MYC regulated, an additional (Figure 1c and Supplementary Table 1). Moreover, the ﬁve genes (FAP, SAT1, COL8A1, LAMA1 and MFAP5) GSEA and the results shown in Figure 5a, together with were similarly affected by both CIP2A and MYC more robust effects of CIP2A depletion on MYC- depletion (Figure 5a). Taken together, these data repressed genes (Figure 5c), clearly indicate that CIP2A demonstrate that 67% (8/12) of the examined CIP2A is predominantly involved in regulation of MYC- signature genes are similarly regulated by both CIP2A mediated gene repression. On the other hand, our data and MYC. As shown in Figure 5c, CIP2A and MYC show that CIP2A effects on MYC protein expression are siRNAs also similarly regulated the expression of four mediated by PP2A activity (Figure 4c). To further prototypical MYC target genes: p15, GADD45A, NCL strengthen the functional links between CIP2A, PP2A and E2F2. Consistent with the observation that the and MYC, we evaluated whether CIP2A’s effects on majority of CIP2A signature genes were induced in regulation of CIP2A- and MYC-repressed genes CIP2A-depleted cells (Figure 1c, Supplementary Table 1), (Figure 5a) were PP2A dependent. As shown in we also found that the MYC-repressed genes (p15 Figure 5e, CIP2A siRNA-elicited induction of all and GADD45A) were more affected than were the CIP2A and MYC repressed genes from Figure 5a, MYC-activated genes (NCL and E2F2) (Figure 5c). except that of COL8A1, was reversed by simultaneous However, CIP2A depletion did signiﬁcantly inhibit the inhibition of either one or both of the PP2A B subunits activity of the MYC/MAX-sensitive minimal E-box B55a and B56b. Interestingly, B-subunit depletion also

Oncogene Identiﬁcation of the CIP2A-regulated transcriptome M Niemela¨ et al 4273 5 4 3 2 1 0 -1 -2 Log2 Fold change -3 -4 -5 siRNA: C M C M C M C M C M C M C M C M C M C M C M C M

FAP SAT1 CAV1 RAB31 PLAUR ACTG2 LAMA1 MFAP5 COL8A1 SLC22A4 RARRES1 SERPINE2

6 CIP2A siRNA 5 MYC siRNA 4 3 2 1 0

Log2 Fold change -1 -2 -3

p15 NCL E2F2 GADD45A

9 p53 8 MYC/E-box 7 NS 6 1.4 * 1.2 5 1 4 0.8 3 0.6 2 0.4 1 0.2 0 β β β β β β α α α α α 0 α Relative luciferase activity siRNA: SCR SCR SCR SCR SCR SCR

siRNA: CIP2A CIP2A CIP2A CIP2A CIP2A CIP2A SCR SCR CIP2A CIP2A CIP2A+B56 CIP2A+B56 CIP2A+B56 CIP2A+B56 CIP2A+B56 CIP2A+B56 CIP2A+B55 CIP2A+B55 CIP2A+B55 CIP2A+B55 CIP2A+B55 CIP2A+B55

FAPSAT1 PLAUR SERPINE2 SLC22A4 COL8A1 Figure 5 CIP2A regulates MYC-mediated transcriptional profile. (a) qRT–PCR quantification of mRNA expression levels of selected CIP2A signature genes in CIP2A (C) or MYC (M) siRNA-treated HeLa cells. Bold indicates genes that have been previously identified as MYC target genes. Mean±s.e.m. log2 fold change as compared with scrambled siRNA-transfected cells in three independent experiments. (b) Heatmap illustrating gene expression levels of CIP2A (KIAA1524), PLAUR, SERPINE2 and SLC22A4 in indicated human breast-derived cell lines. Number of experiments in each cell lines is indicated in parenthesis. Blue indicates downregulation and red upregulation as compared with mean expression levels in total panel of 39 cell lines. (c) qRT–PCR quantification of p15, GADD45, NCL and E2F2 mRNA levels in MYC or CIP2A siRNA-treated HeLa cells. Mean±s.e.m. of three independent experiments is shown. (d) Effects of CIP2A depletion in promoter activities of MYC/E-box and p53-responsive promoter/luciferase reporters. Mean±s.e.m. of two independent experiments (*Po0.05, NS P40.05; two-tailed Student’s t-test) is shown. (e) Effect of depletion of indicated PP2A B subunits on mRNA expression levels of genes that were similarly induced by CIP2A and MYC siRNAs in panel a. Mean±s.e.m. of two independent experiments is shown.

Oncogene Identification of the CIP2A-regulated transcriptome M Niemela¨ et al 4274 reverted MYC-independent inhibition of JNK2 expres- mRNA expression was found in the BRCA1-mutant cell sion by CIP2A siRNA (data not shown). line, HCC1937, whereas the lowest expression was Taken together, these data demonstrate that the detected in the non-cancerous, immortalized mammary CIP2A signature contains validated MYC target genes epithelial cell line, MCF10A (Figure 6d). Importantly, and that CIP2A is predominantly involved in regulation all top seven CIP2A expressing cell lines were of basal- of MYC-repressed genes. Furthermore, it was shown like origin (Figure 6d). We have previously demon- that CIP2A’s effects on gene expression are at large strated that CIP2A promotes MYC expression, ancho- mediated by PP2A. rage-independent proliferation and xenograft tumor growth of basal-like MDA-MB-231 breast cancer cells Clinical relevance of the CIP2A signature in breast cancer (Come et al., 2009). To assess whether CIP2A promotes Both CIP2A and MYC are implicated in human MYC-mediated gene expression in basal-like cells, the breast cancer development (Come et al., 2009; Xu expression of MYC-regulated signature genes (Figure 5a) et al., 2010). Therefore, to assess the clinical relevance in CIP2A siRNA-transfected MDA-MB-231 cells was of the CIP2A-regulated transcriptome, we compared examined. The expression of all of the MYC-regulated the CIP2A signature with two published breast cancer signature genes, except one (LAMA1), was detectable in microarray signatures (Miller et al., 2005; Enerly MDA-MB-231 cells. Furthermore, all expressed genes et al., 2011). By comparing the CIP2A signature with were similarly regulated in CIP2A-depleted MDA-MB- recently published mRNA expression profiles from 231 cells as were observed in both CIP2A- and MYC- 114 human breast cancers (Enerly et al., 2011), we depleted HeLa cells (Figures 6e and 5a). found that the CIP2A signature clustered all cancers Taken together, these results demonstrate a clinical to four major subgroups, out of which the first one association of the CIP2A signature with basal-like and was almost exclusively enriched with both basal-like HER2 þ breast cancers. Moreover, MYC amplifica- and HER2 þ (ERBB2)breastcancersubtypes tion and CIP2A overexpression are not mutually (Figure 6a). Importantly, similar clustering of both exclusive in human breast cancers. Therefore, over- basal-like and HER2 þ subtypes was also found when expression of CIP2A provides a novel explanation for the CIP2A signature was compared with the Miller high MYC protein expression in basal-like and data set (Miller et al., 2005) (Supplementary Figure 7). HER2 þ breast cancers in which the MYC gene The association of the CIP2A signature with both transcription is enhanced either because of increased basal-like and HER2 þ breast cancers suggests that MYC promoter activity or gene amplification (Alles CIP2A might have a profound role in these disease et al., 2009; Chandriani et al., 2009; Xu et al., 2010). subtypes. To further support these conclusions, CIP2A protein expression was evaluated by immunohistochemical staining in a series of 1028 human Discussion breast cancer samples (Joensuu et al., 2003). In accordance with previously published data (Come In this study, we present the first global signaling analysis et al., 2009), CIP2A was found to be expressed in 45% of the function of the PP2A inhibitor protein, CIP2A, in of all breast cancer samples studied (Figure 6b), and human cancer cells. CIP2A is an independently validated its expression correlated with higher tumor grade PP2A inhibitor protein and a clinically relevant human (data not shown). Moreover, in accordance with the oncoprotein. Therefore, it was hypothesized that character- data shown in Figure 6a, CIP2A was overexpressed ization of the CIP2A-regulated transcriptome would reveal significantly more frequently in basal-like and novel insights into the signaling mechanisms downstream HER2 þ breast cancers than in all breast tumors of oncogenic PP2A inhibition. To circumvent the acknowl- analyzed (Figure 6b and Supplementary Figure 8A). edged limitations in the repeatability of published micro- To assess whether CIP2A expression and MYC array results (Ioannidis et al.,2009),thereportedCIP2A amplifications associate, the same breast tumor material signature consists only of genes regulated by CIP2A was analyzed for MYC amplifications by fluorescence depletion in all five independent sample pairs. All tested in situ hybridization (FISH) assay. On the basis of a recent signature genes were shown by quantitative RT–PCR to be meta-analysis of 29 studies, it was concluded that the regulated by CIP2A in four cancer cell lines (Figures 1d average frequency of MYC amplification in breast tumors, and 6e and Supplementary Figures 1B and C) in which which was defined as a two-fold increase in gene copy CIP2A has previously been shown to be overexpressed number, was 15.7% (Deming et al., 2000; Xu et al., 2010). (Junttila et al., 2007; Come et al., 2009; Khanna et al., 2009; In our material of 144 representative samples, the Choi et al., 2011). Thus, we consider the reported CIP2A frequency of MYC amplification was 18.1% (Figure 6c) signature to be a reliable representation of the transcrip- (a representative image of MYC amplification is shown in tional program driven by CIP2A in human cancer cells. Supplementary Figure 8B). Interestingly, rather than being Pathway-level analysis of the signature genes indi- mutually exclusive, MYC amplification significantly asso- cated that CIP2A regulates both MYC-independent and ciated with CIP2A expression (Figure 6c). MYC-dependent signaling. With regard to MYC- The association of CIP2A with basal-like breast independent CIP2A functions, we discovered a novel cancers also became evident from the analysis of CIP2A role for CIP2A in promoting JNK2 expression and mRNA expression levels across 40 breast cancer cell transwell migration in HeLa cells, whereas MYC lines. Out of all cell lines examined, the highest CIP2A depletion did not affect either of these phenotypes

Oncogene Identiﬁcation of the CIP2A-regulated transcriptome M Niemela¨ et al 4275

Normal Lum A Lum B Basal ERBB2

CIP2A IHC MYC ampliﬁcation Negative Positive Neg. Pos. Total All tumors 569 (55%) 459 (45%) 0 65 (95.6%) 3 (4.4%) 68 Basal-like 34 (39%) 53 (61%) p= 0.0014 CIP2A 1 32 (68.1%) 15 (31.9%) 47 HER2+ 76 (40%) 116 (60%) p<0.0001 2 21 (72.4%) 8 (27.6%) 29

Total 118 (81.9%) 26 (18.1%) 144

p<0.001 Basal

MDA-MB-231

3 CIP2A siRNA 2

-1 Log2 Fold change

-2

FAP SAT1

PLAUR MFAP5 COL8A1 SLC22A4 SERPINE2

Figure 6 The CIP2A signature associates with basal-type breast cancer. (a) Hierarchical clustering of different human breast cancer subtypes based on the CIP2A signature. Breast cancer subtype gene expression profiles were acquired from the study by Enerly et al. (2011). The classifier separated the 114 breast tumors based on the subtypes of breast cancer. (b) Immunohistochemical analysis of CIP2A expression in human breast cancer tissue microarray material revealed statistically significant associations between CIP2A expression and basal-like and HER2 þ subtypes in comparison with other tumors (all tumors (n ¼ 1028); basal-like (n ¼ 87) and HER2 þ (n ¼ 192). (c) Significant association of MYC gene amplification and CIP2A expression in human breast cancer tissue microarray material. (d) Expression of CIP2A mRNA across a panel of 39 breast cancer cell lines. The box plots are sorted in descending order starting with the cell lines showing the highest expression of CIP2A (e) qRT–PCR analysis of mRNA expression levels of indicated signature genes in CIP2A siRNA-treated MDA-MD-231 cells. Genes included in the assay were selected based on similar responsiveness to both MYC and CIP2A siRNAs in HeLa cells in Figure 5a. Bold indicates genes that have been previously identified as MYC target genes. Mean±s.e.m. log2 fold change as compared with scrambles siRNA-transfected cells in three independent experiments.

(Figure 3). Moreover, among the RT–PCR-validated dent genes included in the CIP2A signature will reveal signature genes, we identiﬁed genes that were regulated novel mechanisms by which CIP2A regulates by CIP2A, but not by MYC depletion (Figure 5a). cell behavior independently of its function in MYC It is anticipated that further analysis of MYC-indepen- stabilization.

Oncogene Identification of the CIP2A-regulated transcriptome M Niemela¨ et al 4276 Regardless of the established role for CIP2A in In conclusion, we have identified and validated a regulating MYC expression (Junttila et al.,2007;Come CIP2A-driven transcriptional signature. Pathway ana- et al., 2009; Dong et al., 2010), thus far, the contribution of lysis of these signature genes revealed both MYC- CIP2AtoMYC-mediatedgeneexpressionregulationhas dependent and MYC-independent functions of CIP2A. not been analyzed. Here, pathway analysis of the CIP2A Importantly, our data both validate the role for CIP2A signature genes revealed significant association of the day- in regulating established MYC target genes and identify 5 CIP2A signature with MYC-associated signaling (Fig- novel MYC-regulated genes. We also provide the first ure 2). Moreover, RT–PCR analysis verified that CIP2A demonstration that CIP2A’s effects on cancer cell depletion significantly modulated the expression of four behavior are mediated by PP2A complexes. Impor- direct MYC target genes (SLC22A4, SERPINE2, p15 and tantly, the clinical significance of this study is high- GADD45A). In addition, we identified several novel lighted by a signature-associated identification of the MYC-regulated genes from the CIP2A signature linkage between basal-like and HER2 þ breast cancers (Figure 5a). Taken together, the results of this study and CIP2A. Taken together, it is expected that the validate the functional role of CIP2A in regulating MYC- identified signature will be an important resource for mediated cellular signaling output. understanding how the emerging human oncoprotein Transcriptional signatures are being used to classify CIP2A exerts its oncogenic functions in human malig- cancer subtypes and for patient stratification in cancer nancies. Finally, it is tempting to speculate that this therapy trials. Here, we show that among the different signature may help in future patient stratification breast cancer subtypes, the CIP2A-regulated transcrip- methods for current adjuvant therapies and in future tome clustered most clearly together with basal-like and for MYC- and/or CIP2A-targeted therapies. HER2 þ breast tumors (Figure 6a). Moreover, the CIP2A protein was found to be predominantly overexpressed in both of these breast cancer subtypes Materials and methods (Figure 6b). On the other hand, in breast cancer cell lines, the highest CIP2A expression levels were detected Microarray and RT–qPCR in cells of basal-like origin (Figure 6d). Taken together, RNA was extracted using NucleoSpin RNA II kit (Macherey- Nagel, Du¨ren, Germany) according to the manufacturer’s these results strongly indicate that CIP2A is closely instructions. Reverse transcription was performed in the presence linked to the basal-like breast cancer subtype and is also of RNase inhibitor rRNAsin (Promega, Madison, WI, USA) involved in HER2 þ cancers. On the basis of our using M-MuLV RNase H- reverse transcriptase (Finnzymes, results, it is clear that further verification of the value of ThermoFisher, Waltham, MA, USA). Genome-wide gene expres- CIP2A signature genes and CIP2A itself, in terms of sion study was performed using the Illumina HT-12 v.3 more a precise classification and treatment stratification microarray (Illumina, San Diego, CA, USA). The microarray of breast cancer patients, deserves further attention. data were extracted using Illumina BeadStudio software and It has been suggested that post-translational MYC analyzed using the lumi Bioconductor package (Bioconductor, protein stabilization may have an important role in Seattle, WA, USA). The data were preprocessed with the variance MYC-mediated oncogenic functions in breast cancer (Xu stabilization transformation algorithm (Lin et al., 2008) followed by quantile normalization (Bolstad et al., 2003). Probes with et al., 2010). However, the mechanisms by which MYC expression values below the background level (detection P-value might be stabilized in basal-like and HER2 þ breast o0.05) in all samples were discarded from further analyses to cancers have not been identified as yet. The strong reduce the number of false positives. Genes with an absolute fold expression of the MYC-stabilizing protein CIP2A and change of 1.3 were considered as upregulated or downregulated. clinical association of its signature in basal-like and RT–qPCR-assays were designed using the Universal ProbeLibrary HER2 þ breast cancers identified in this study provides a design tool (Roche Applied Science, Indianapolis, IN, USA). novel, plausible explanation for the increased MYC protein Primer sequences and amplicon sizes are presented in Supplemen- levels and activity in these breast cancer subtypes. tary Table 3. RT–PCR reactions were run using Applied Association of MYC amplification and CIP2A expression Biosystems 7900HT Fast Sequence Detection System and Taq- in breast cancer is another novel finding of this study and Man Universal Master Mix II, no UNG (Applied Biosystems, Life Technologies, Carlsbad, CA, USA). The data were normal- suggests that inhibition of bidirectional positive feedback ized using b-actin as a reference gene. PP2A B-subunit knockdown loop between MYC and CIP2A (Khanna et al., 2009; efficiency was measured using previously published RT–qPCR Mannava et al., 2011) could be an attractive therapeutic primers (Sablina et al., 2010) using DyNAmo HS SYBR Green target in breast cancers in which these two oncogenic qPCR Kit (Finnzymes). alterations exist. Moreover, it is plausible that further functional analysis of MYC-dependent CIP2A signature siRNA transfections and antibodies genes will reveal novel mechanisms by which these two HeLa, MDA-MB-231, MCF-7 and AGS cells on a 6-well plate oncoproteins together promote disease progression in were transfected with 250 pmol of siRNA with oligofectamine basal-like and HER2 þ breast cancers. Furthermore, our (Invitrogen) according to the manufacturer’s instructions. siRNA current results, together with the demonstrated tumor sequences are presented in Supplementary Table 3. Antibodies: CIP2A for immunoblots: (2G10-3B5) sc80659 (Santa Cruz growth-promoting activity of CIP2A in MDA-MB-231 Biotechnology Inc., Santa Cruz, CA, USA), CIP2A for IHC: cells (Come et al., 2009), indicate that either CIP2A or rabbit polyclonal antibody (Soo Hoo et al., 2002), b-actin (AC- CIP2A signature genes identified herein present as attractive 74) A5316 (Sigma-Aldrich, St Louis, MO, USA), c-myc (9E10) novel targets for cancer therapeutics in these breast cancer (MMS150R) (Nordic Biosite, Ta¨by, Sweden), JNK-2 (D-2) sc- subtypes. 7345 (Santa Cruz Biotechnology Inc.).

Oncogene Identification of the CIP2A-regulated transcriptome M Niemela¨ et al 4277 Bioinformatic analysis Immunohistochemistry Identification of signaling networks and biological functions Before immunostaining of CIP2A, tissue microarray sections associated with CIP2A signature genes was conducted by were deparaffinized and rehydrated, and endogenous peroxidase ingenuity pathway analysis (Ingenuity Systems, Redwood activity in tissues was blocked using 3% hydrogen peroxide. City, CA, USA) according to their standard procedures. Antigen retrieval was carried out in Tris–EDTA buffer (pH 9.0, To identify association of the CIP2A signature with human 10 mM Tris, 1 mM EDTA) by heating sections in water bath breast cancer subtypes, we extracted preprocessed publicly (98 1C, 30 min). The CIP2A antibody was diluted in PowerVision available breast cancer gene expression profiling data sets with blocking solution (1:5000), and incubated on sections overnight breast cancer molecular subtype annotation from the National at 4 1C. The primary antibody was detected using Power- Center for Biotechnology Information (NCBI) Gene Expres- Vision þ Poly-HRP-histostaining kit (DVPB þ 110DAP, Tech- sion Omnibus. These data sets can be accessed through series nologies Co., Daly City, CA, USA) following the manufacturer’s accession numbers GSE19783, GSE4922 and GSE1456. protocol. CIP2A expression was graded based on cytoplasmic GSE19783 is based on Agilent microarray platform (Agilent, staining intensity as negative or positive. Immunostainings and Santa Clara, CA, USA), whereas GSE4922 and GSE1456 are gradings for ER, PgR, HER2, EGFR and CK5 were performed both based on Affymetrix microarray platforms (Affymetrix, as described in detail earlier (Sihto et al., 2008). Cancers were Santa Clara, CA, USA; U133 A and B arrays). Owing to the grouped into five molecular subtype based on their expression complexity in combining data from two different microarray profile (Luminal A: ER þ and/or PgR þ ,HER2À;LuminalB: platforms, we decided to perform the analysis in each platform ER þ and/or PgR þ ,HER2þ ; HER2: ERÀ and PgRÀ, type separately. U133A and U133B chips were preprocessed HER2 þ ; Basal-like: ERÀ,PgRÀ,HER2À,CK5þ and/or separately using the R language (R Development core team, EGFR þ : Non-expressor: all five markers negative). Vienna, Austria) and the RMA method implemented in the Bioconductor (Gentleman et al., 2004) package affy. Alter- FISH analysis native CDF files mapping Affymetrix probes directly to The MYC copy number was assessed from 436 out of 1028 Ensembl gene ids were used in preprocessing (Dai et al., breast cancers using FISH. A total of 292 samples were 2005). Data from U133A and U133B were combined by rejected from analysis because of missing tumor tissue on calculating the medians of genes appearing on both array tissue microarray slides or poor FISH quality, leaving total of types. The tumors were classified into different breast cancer 144 samples into series. Samples were hybridized with Vysis subtypes as described previously in the original articles (Miller LSI C-MYC (8q24.12-q24.13) SpectrumOrange Probe and et al., 2005; Enerly et al., 2011). We performed hierarchical centromeric CEP 8 (D8Z2) probe according to the manufac- clustering of the 134 CIP2A signature genes across 114 human turer’s instructions (Abbot Molecular, Des Plaines, IL, USA). breast tumors from GSE19783 and across 251 breast tumors Ratio of orange-labeled MYC and green-labeled centromere 8 from GSE4922 and GSE1456. FISH signals were assessed within tumor cells using Â 1000 To analyze CIP2A expression in human breast-derived cell magnification. A signal ratio two or more were considered to lines, gene expression data of 39 human breast cancer cell lines represent MYC amplification. If 410% of tumor cells were were extracted from the National Cancer Institute’s cancer estimated to contain MYC amplification, MYC/centromere 8 Bioinformatics Grid (caBIG). The data were pre-processed signal ratio was assessed only in cells with amplification. using R and RMA method implemented aroma.affymetrix Samples were analyzed using Zeiss Axioplan 2 fluorescence package. The data were originally generated using Affymetrix microscope and images were acquired with AxioVision GeneChip U133 Plus 2.0 arrays. On the basis of these data, a software (Carl Zeiss MicroImaging GmbH, Jena, Germany). heatmap of the expression difference of CIP2A (KIAA1524), PLAUR, SERPINE2 and SLC22A4 across all the breast cancer cell lines was also generated. Conflict of interest

Breast cancer patient material The authors declare no conflict of interest. Nationwide population-based breast cancer series was constructed as described in detail elsewhere (Joensuu et al., 2003). Inclusion criteria for subjects based on clinical data Acknowledgements are presented in Supplementary Materials and Methods. Representative tumor samples for CIP2A immunohisto- We thank Taina Kalevo-Mattila for expert technical assistance. chemistry were available for 1228 subjects, in whom molecular Melissa R Junttila is acknowledged for PP2A assay data and breast cancer subtypes were successfully defined for 1028 Minna-Maija Lintunen for performing the FISH hybridizations. tumors. These 1028 tumors were entered into the study. Panu Jaakkola is acknowledged for B-subunit siRNAs. The Formalin-fixed, paraffin-embedded tumor samples were col- Finnish DNA Microarray centre is acknowledged for the lected from hospital archives. Representative tumor regions microarray and qRT–PCR analysis. This study was supported were defined from tumor samples using hematoxylin and eosin by grants from the Academy of Finland (grant no. 125826, staining, and tissue microarray blocks were constructed from 1131449 and 8217676), Sigrid Juselius Foundation, the Cancer those tumors using 0.6 mm diameter needle. Permission to use Society of Finland, Helsinki University Central Hospital Research tissues for research purposes was provided by the Ministry of Funds (TYH2009304) and the Foundation of Finnish Cancer Social Affairs and Health, Finland (permission 123/08/97). Institute.

References

Alles MC, Gardiner-Garden M, Nott DJ, Wang Y, Foekens JA, relative to ER status reveal elevated activity of MYC and E2F in the Sutherland RL et al. (2009). Meta-analysis and gene set enrichment ‘basal’ breast cancer subgroup. PLoS One 4: e4710.

Oncogene Identification of the CIP2A-regulated transcriptome M Niemela¨ et al 4278 Arnold HK, Sears RC. (2006). Protein phosphatase 2A regulatory Junttila MR, Puustinen P, Niemela M, Ahola R, Arnold H, Bottzauw subunit B56alpha associates with c-myc and negatively regulates T et al. (2007). CIP2A inhibits PP2A in human malignancies. Cell c-myc accumulation. Mol Cell Biol 26: 2832–2844. 130: 51–62. Ben-Israel H, Sharf R, Rechavi G, Kleinberger T. (2008). Adenovirus Katz J, Jakymiw A, Ducksworth MK, Stewart CM, Bhattacharyya I, E4orf4 protein downregulates MYC expression through interaction Cha S et al. (2010). CIP2A expression and localization in oral with the PP2A-B55 subunit. J Virol 82: 9381–9388. carcinoma and dysplasia. Cancer Biol Ther 10: 694–699. Bolstad BM, Irizarry RA, Astrand M, Speed TP. (2003). A comparison Khanna A, Bockelman C, Hemmes A, Junttila MR, Wiksten JP, of normalization methods for high density oligonucleotide array data Lundin M et al. (2009). MYC-dependent regulation and based on variance and bias. Bioinformatics 19: 185–193. prognostic role of CIP2A in gastric cancer. J Natl Cancer Inst Cappellen D, Schlange T, Bauer M, Maurer F, Hynes NE. (2007). 101: 793–805. Novel c-MYC target genes mediate differential effects on cell Li W, Ge Z, Liu C, Liu Z, Bjorkholm M, Jia J et al. (2008). CIP2A is proliferation and migration. EMBO Rep 8: 70–76. overexpressed in gastric cancer and its depletion leads to impaired Chandriani S, Frengen E, Cowling VH, Pendergrass SA, Perou CM, clonogenicity, senescence, or differentiation of tumor cells. Clin Whitfield ML et al. (2009). A core MYC gene expression signature Cancer Res 14: 3722–3728. is prominent in basal-like breast cancer but only partially overlaps Li Z, Van Calcar S, Qu C, Cavenee WK, Zhang MQ, Ren B. (2003). A the core serum response. PLoS One 4: e6693. global transcriptional regulatory role for c-Myc in Burkitt’s Chen KF, Liu CY, Lin YC, Yu HC, Liu TH, Hou DR et al. (2010). lymphoma cells. Proc Natl Acad Sci USA 100: 8164–8169. CIP2A mediates effects of bortezomib on phospho-Akt and Lin SM, Du P, Huber W, Kibbe WA. (2008). Model-based variance- apoptosis in hepatocellular carcinoma cells. Oncogene 29: 6257–6266. stabilizing transformation for Illumina microarray data. Nucleic Chen W, Possemato R, Campbell KT, Plattner CA, Pallas DC, Hahn Acids Res 36: e11. WC. (2004). Identification of specific PP2A complexes involved in Mathiasen DP, Egebjerg C, Andersen SH, Rafn B, Puustinen P, human cell transformation. Cancer Cell 5: 127–136. Khanna A et al. (2011). Identification of a c-Jun N-terminal kinase- Choi YA, Park JS, Park MY, Oh KS, Lee MS, Lim JS et al. (2011). 2-dependent signal amplification cascade that regulates c-Myc levels Increase in CIP2A expression is associated with doxorubicin in ras transformation. Oncogene (e-pub ahead of print 27 June 2011; resistance. FEBS Lett 585: 755–760. doi:10.1038/onc.2011.230). Come C, Laine A, Chanrion M, Edgren H, Mattila E, Liu X et al. Mannava S, Omilian AR, Wawrzyniak JA, Fink EE, Zhuang D, (2009). CIP2A is associated with human breast cancer aggressivity. Miecznikowski JC et al. (2011). PP2A-B56a controls oncogene- Clin Cancer Res 15: 5092–5100. induced senescence in normal and tumor human melanocytic cells. Dai M, Wang P, Boyd AD, Kostov G, Athey B, Jones EG et al. (2005). Oncogene (e-pub ahead of print 8 August 2011; doi:10.1038/ Evolving gene/transcript definitions significantly alter the inter- onc.2011.339). pretation of GeneChip data. Nucleic Acids Res 33: e175. Meyer N, Penn LZ. (2008). Reflecting on 25 years with MYC. Nat Rev Deming SL, Nass SJ, Dickson RB, Trock BJ. (2000). C-myc Cancer 8: 976–990. amplification in breast cancer: a meta-analysis of its occurrence Miller LD, Smeds J, George J, Vega VB, Vergara L, Ploner A et al. and prognostic relevance. Br J Cancer 83: 1688–1695. (2005). An expression signature for p53 status in human breast Dong QZ, Wang Y, Dong XJ, Li ZX, Tang ZP, Cui QZ et al. (2010). cancer predicts mutation status, transcriptional effects, and patient CIP2A is overexpressed in non-small cell lung cancer and correlates survival. Proc Natl Acad Sci USA 102: 13550–13555. with poor prognosis. Ann Surg Oncol 18: 857–865. Mumby M. (2007). PP2A: unveiling a reluctant tumor suppressor. Cell Eichhorn PJ, Creyghton MP, Bernards R. (2009). Protein phosphatase 130: 21–24. 2A regulatory subunits and cancer. Biochim Biophys Acta 1795: 1–15. Qu W, Li W, Wei L, Xing L, Wang X, Yu J. (2010). CIP2A is Enerly E, Steinfeld I, Kleivi K, Leivonen SK, Aure MR, Russnes HG overexpressed in esophageal squamous cell carcinoma. Med Oncol et al. (2011). miRNA-mRNA integrated analysis reveals roles for (e-pub ahead of print). miRNAs in primary breast tumors. PLoS One 6: e16915. Rangarajan A, Hong SJ, Gifford A, Weinberg RA. (2004). Species- Fernandez PC, Frank SR, Wang L, Schroeder M, Liu S, Greene J et al. and cell type-specific requirements for cellular transformation. (2003). Genomic targets of the human c-Myc protein. Genes Dev 17: Cancer Cell 6: 171–183. 1115–1129. Sablina AA, Hector M, Colpaert N, Hahn WC. (2010). Identification Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S of PP2A complexes and pathways involved in cell transformation. et al. (2004). Bioconductor: open software development for Cancer Res 70: 10474–10484. computational biology and bioinformatics. Genome Biol 5: R80. Sihto H, Lundin J, Lehtimaki T, Sarlomo-Rikala M, Butzow R, Holli Guenebeaud C, Goldschneider D, Castets M, Guix C, Chazot G, K et al. (2008). Molecular subtypes of breast cancers detected in Delloye-Bourgeois C et al. (2010). The dependence receptor mammography screening and outside of screening. Clin Cancer Res UNC5H2/B triggers apoptosis via PP2A-mediated dephosphoryla- 14: 4103–4110. tion of DAP kinase. Mol Cell 40: 863–876. Soo Hoo L, Zhang JY, Chan EK. (2002). Cloning and characterization Hahn WC, Counter CM, Lundberg AS, Beijersbergen RL, Brooks of a novel 90 kDa ‘companion, auto-antigen of p62 overexpressed in MW, Weinberg RA. (1999). Creation of human tumour cells with cancer. Oncogene 21: 5006–5015. defined genetic elements. Nature 400: 464–468. Vaarala MH, Vaisanen MR, Ristimaki A. (2010). CIP2A expression is Ioannidis JP, Allison DB, Ball CA, Coulibaly I, Cui X, Culhane AC increased in prostate cancer. J Exp Clin Cancer Res 29: 136. et al. (2009). Repeatability of published microarray gene expression Virshup DM, Shenolikar S. (2009). From promiscuity to precision: analyses. Nat Genet 41: 149–155. protein phosphatases get a makeover. Mol Cell 33: 537–545. Janssens V, Longin S, Goris J. (2008). PP2A holoenzyme assembly: Westermarck J, Hahn WC. (2008). Multiple pathways regulated by the in cauda venenum (the sting is in the tail). Trends Biochem Sci 33: tumor suppressor PP2A in transformation. Trends Mol Med 14: 113–121. 152–160. Joensuu H, Isola J, Lundin M, Salminen T, Holli K, Kataja V et al. Xu J, Chen Y, Olopade OI. (2010). MYC and breast cancer. Genes (2003). Amplification of erbB2 and erbB2 expression are superior to Cancer 1: 629–640. estrogen receptor status as risk factors for distant recurrence in Zhao JJ, Roberts TM, Hahn WC. (2004). Functional genetics and pT1N0M0 breast cancer: a nationwide population-based study. Clin experimental models of human cancer. Trends Mol Med 10: Cancer Res 9: 923–930. 344–350.

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

Oncogene