Oncogene (2012) 31, 1299–1310 & 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc ORIGINAL ARTICLE HES6 is selectively overexpressed in glioma and represents an important transcriptional regulator of glioma proliferation

S Haapa-Paananen1, S Kiviluoto1, M Waltari2, M Puputti2, JP Mpindi1,3, P Kohonen1, O Tynninen4, H Haapasalo5, H Joensuu2, M Pera¨la¨1 and O Kallioniemi1,3

1Department of Medical Biotechnology, VTT Technical Research Centre of Finland and Centre for Biotechnology, University of Turku, Turku, Finland; 2Department of Oncology, University of Helsinki and Helsinki University Central Hospital, Helsinki, Finland; 3FIMM—Institute for Molecular Medicine Finland, University of Helsinki, Helsinki, Finland; 4Department of Pathology, Helsinki University Central Hospital and Haartman Institute, University of Helsinki, Helsinki, Finland and 5Department of Pathology, Tampere University Hospital, Tampere, Finland

Malignant glioma is the most common brain tumor with Introduction 16 000 new cases diagnosed annually in the United States. We performed a systematic large-scale transcriptomics Malignant glioma is the most common brain cancer with data mining study of 9783 tissue samples from the B16 000 new cases diagnosed annually in the United GeneSapiens database to systematically identify States. High-grade malignant gliomas comprise of the that are most glioma-specific. We searched for genes that World Health Organization (WHO) grade IV brain were highly expressed in 322 glioblastoma multiforme tumors, including glioblastoma multiforme (GBM) and tissue samples and 66 anaplastic astrocytomas as its variants, and the WHO grade III tumors (anaplastic compared with 425 samples from histologically normal forms of astrocytoma, oligodendroglioma and oligoas- . cofactor HES6 trocytoma). GBM is the most common primary brain (hairy and of split 6) emerged as the most tumor in adults. Primary GBM forms de novo without a glioma-specific gene. Immunostaining of a tissue micro- previous history of a low-grade disease, whereas array showed HES6 expression in 335 (98.8%) out of the secondary GBM progresses from previously diagnosed 339 glioma samples. HES6 was expressed in endothelial low-grade glioma. Patients with newly diagnosed GBM cells of the normal brain and glioma tissue. Recurrent have a median survival of B1 year, and GBMs usually grade 2 astrocytomas and grade 2 or 3 oligodendroglio- respond relatively poorly to and radio- mas showed higher levels of HES6 immunoreactivity than therapy (Ohgaki and Kleihues, 2007, 2009; Kumar et al., the corresponding primary tumors. High HES6 mRNA 2008). Tumor recurrence is common after current expression correlated with the proneural subtype that treatments, and the disease follows a fatal course in generally has a favorable outcome but is prone to recur. virtually all patients. Functional studies suggested an important role for HES6 Recently, The Cancer Genome Atlas (TCGA) Project in supporting survival of glioma cells, as evidenced by was launched to comprehensively profile 500 GBM reduction of cancer cell proliferation and migration after samples on several microarray platforms, such as gene HES6 silencing. The biological role and consequences of , single-nucleotide polymorphism, copy number HES6 silencing and overexpression was explored with and expression microarrays (Cancer Genome Atlas genome-wide analyses, which implicated a role for HES6 Research Network, 2008; Noushmehr et al., 2010; in , c- and nuclear factor-jB transcriptional Verhaak et al., 2010). A number of other microarray networks. We conclude that HES6 is important expression profiling studies on gliomas have also been for glioma cell proliferation and migration, and may have published. These data sets enable the development a role in angiogenesis. of systems biological knowledge on the molecular and Oncogene (2012) 31, 1299–1310; doi:10.1038/onc.2011.316; genetic regulatory networks in glioma. published online 25 July 2011 In this study, we mined for glioma-specific genes from our GeneSapiens database that contains Keywords: HES6; glioma; astrocytoma; glioblastoma; data from 9783 samples covering 175 types of healthy TMA; RNAi and pathological tissues (Kilpinen et al., 2008). This database contains Affymetrix (Santa Clara, CA, USA) gene expression array experiments and makes it possible to compare mRNA expression levels of genes across Correspondence: Dr S Haapa-Paananen, Department of Medical different tissues. The database includes data from 425 Biotechnology, VTT Technical Research Centre of Finland and Centre histologically normal tissue samples from the central for Biotechnology, Ita¨inen Pitka¨katu 4C, PO Box 106, FI-20521 nervous system and 475 glioma samples (322 GBMs, 66 Turku, Finland. E-mail: saija.haapa-paananen@vtt.fi anaplastic astrocytomas, 10 astrocytomas, 44 anaplastic Received 20 February 2011; revised 14 June 2011; accepted 21 June 2011; oligodendrogliomas, 15 oligodendrogliomas and 18 published online 25 July 2011 mixed gliomas). Here, we systemically sought for HES6 in glioma S Haapa-Paananen et al 1300 putative mRNA expression-based biomarkers in GBM was associated with proneural tumors, whereas me- and anaplastic astrocytoma and found that HES6 (hairy senchymal tumor types had the lowest expression levels. and enhancer of split 6) was highly expressed in gliomas as compared with the normal brain tissues. HES6 expression in tumors HES6 is a member of the basic helix-loop-helix family HES6 protein levels were studied in a glioma tissue of transcription factors. Its overexpression has been microarray (TMA) with 417 samples and in 9 normal reported in metastatic colon and prostate cancers brain tissue samples. Tumor HES6 expression could be (Swearingen et al., 2003; Vias et al., 2008), and in assessed in 339 (81.3%) out of the 417 cases. HES6 cancers of the , breast and kidney (Swearingen expression was detected in 335 (98.8%) out of the 339 et al., 2003; Hartman et al., 2009). Hes6 inhibits Hes1 tumors regardless of glioma tumor histology (Table 1). A activity by forming heterodimers with Hes1, supports total of 74 (21.8%) of the tumors showed low levels of achaete-scute complex homolog-1 (Ascl1) activity and HES6 expression ( þ ), 83 (24.3%) moderate ( þþ)and promotes neuronal differentiation (Bae et al., 2000; 178 (52.5%) strong expression ( þþþ). Six out of the Koyano-Nakagawa et al., 2000). During cortical neu- nine normal brain tissue samples were negative and three rogenesis and in the presence of astrocyte inducers showed low HES6 staining. We also analyzed 30 non- ciliary neurotrophic factor or leukemia inhibitory glioma brain tumors, including medulloblastomas, me- factor, HES6 can prevent the precursor-to-astrocyte ningiomas and schwannomas. Of these, 18 (60%) showed transition and promote neuronal differentiation through negative (À), 9 (30%) low ( þ ) and 3 (10%) moderate independent mechanisms (Jhas et al., 2006). Progressive ( þþ) HES6 expression; none showed strong expression. astrocytoma and secondary GBM formation are char- HES6 immunoreactivity in gliomas was mainly acterized by inhibition of Notch signaling; upregulation nuclear (Figure 2). Low-intensity cytoplasmic staining of DLL1, ASCL1 and HES6, and uninduced levels of was seen in astrocytomas with gemistocytic cells HES1 (Somasundaram et al., 2005). Here, we evaluated (Figure 2e). There was no statistical association between the role of HES6 as a potential biomarker in glioma and HES6 staining intensity, the histological tumor type or its functional importance. the WHO grade. HES6 staining of endothelial cells was present in 75% of glioma samples and in four out of nine samples of non-neoplastic cerebral white matter and cortex (arrows in Figure 2). Results HES6 protein was expressed more strongly in tissue biopsies taken from recurrent brain tumors as compared Systematic body-wide search for overexpressed with tissue collected at the time of first diagnosis genes in glioma from the GeneSapiens database (Table 1). HES6 was expressed strongly ( þþþ) We searched for potential biomarkers in anaplastic in 12 (80%) out of the 15 recurrent astrocytomas as astrocytoma and GBM as compared with normal brain compared with 19 (54%) out of the 35 primary tissue on the basis of the genome-wide transcriptomics astrocytomas (P ¼ 0.039). Similarly, 11 (73%) out data in the GeneSapiens database (Kilpinen et al., 2008). of the 15 recurrent oligodendrogliomas expressed The expression of the identified genes was then HES6 strongly as compared with only 5 (29%) of the evaluated across all normal and cancerous tissues. The 17 primary oligodendrogliomas (P ¼ 0.013). In GBMs, genome- and body-wide mining implicated transcription no significant difference in HES6 expression was present cofactor HES6 as the most highly ranked gene between primary versus recurrent or primary versus in anaplastic astrocytoma, and with a rank number 10 secondary GBM tumors (P ¼ 0.745 and P ¼ 0.215, in GBM. HES6 gene was highly expressed in gliomas respectively). as compared with both normal brain tissues, as well as with all other normal and cancer tissues (Figures 1a–c). Associations between HES6 protein expression with High HES6 expression was present in anaplastic PDGFRA, KIT, VEGFR2 and EGFR gene copy astrocytoma, GBM, oligodendroglioma and mixed numbers and protein levels glioma (Figure 1c). High HES6 mRNA expression in Platelet-derived growth factor alpha (PDGFRA), these tumor types (n ¼ 452) was also observed in the KIT, vascular endothelial growth factor receptor 2 REpository for Molecular BRAin Neoplasia DaTa (VEGFR2)andEGFR gene copy numbers were deter- (Rembrandt) (Madhavan et al., 2009) when compared mined in a subset of patients (n ¼ 37) diagnosed with with the normal brain tissues (n ¼ 28; Supplementary primary GBM using chromogenic in-situ hybridization File S1A). Neuroblastoma, melanoma, colon cancer and (CISH) and compared with tumor HES6 expression. some breast cancers displayed slightly elevated expres- Presence of strong ( þþþ) tumor HES6 expression was sion of HES6 (Figure 1b). In comparison, a body-wide compared with less expression (À, þ or þþ). expression plot for epidermal growth factor receptor Amplifications of PDGFRA, KIT and VEGFR2,which (EGFR; Supplementary File S1B) showed high expres- are adjacent genes on 4q, were associated sion in glioma, as well as cancers of the lung, head and with strong tumor HES6 expression (Supplementary File neck, thyroid, bladder and mesothelioma, in addition to S2A). The association between amplified PDGFRA and several normal tissues. Out of the glioblastoma samples HES6 expression was particularly strong. PDGFRA available in both GeneSapiens (Figure 1d) and TCGA amplification was detected in 5 (19.2%) out of the 26 (Supplementary File S1C), high HES6 gene expression GBMs successfully analyzed, and 4 (80%) out of the 5

Oncogene HES6 in glioma S Haapa-Paananen et al 1301 HES6 Healthy Cancer Other Diseases HES6 blood unspecified glioma brain 1500 leukocyt. lymphatic system nucleus accumbens frontal lobe parietal lobe temporal lobe 1000 occipital lobe hippocampus putamen amygdala 500 subthalamic nucleus hypothalamus brain stem midbrain sustantia nigra Relative Expression Level superior vestibular nuclei 0 medulla oblongata cerebellum 2000 4000 6000 8000 10000 corpus callosum spinal cord Samples spinal ganglion glioblastoma multiforme anaplastic oligodendroglioma Anatomical system anaplastic astrocytoma Hematological mixed glioma GI tract & organs Connectivity and muscular Urogenital 0 500 1000 1500 Respiratory Gynegological & Breast Nervous Relative Expression Level Endocrine & Salivary Stem cells

HES6 HES6 1500 1500

1000 1000

500 500 0

Relative Expression Level GBM GBM GBM GBM GBM Relative Expression Level 0 Proneural ungrouped Classical Neural Mesenchymal GIST Glioma Melanoma cancer Liposarcoma Renal cancer Sarcoma, NOS Bladder cancer Neuroblastoma Uterine sarcoma Chondrosarcoma Testis, seminoma Pancreatic cancer Thyroid carcinoma Renal, oncocytoma Breast ductal cancer Colorectal carcinoma Breast lobular cancer Breast cancer, others Lung, carcinoid tumor Ovarian tumor, others Lung adenocarcinoma Lung, large cell cancer Breast carcinoma, NOS Uterine, Mullerian tumor Vagina/Vulva carcinoma Gastric adenocarcinoma Uterine adenocarcinoma Cervical adenocarcinoma Prostate adenocarcinoma Ovarian, serous carcinoma Esophagus adenocarcinoma Peritoneum adenocarcinoma Ovarian, clear cell carcinoma Ovarian, mucinous carcinoma Ovarian adenocarcinoma, NOS Lung, squamous cell carcinoma Small intestine, adenocarcinoma Ovarian, endometrioid carcinoma Uterine squamous cell carcinoma Cervical squamous cell carcinoma Figure 1 (a) Body-wide expression profile of HES6 across the GeneSapiens database. Each dot represents gene expression in one sample. Anatomical origins of each sample are marked with colored bars below the gene plot (legend at the bottom left corner). Sample types with higher than average expression or an outlier expression profile are shown colored in the figure legend at the top left corner. (b) HES6 expression box-plot in more detailed cancer types. The box refers to the quartile distribution (25–75%) range, with the median shown as a red horizontal line. In addition, the 95% black whiskers and individual outlier samples are shown. (c) HES6 expression box-plot in central nervous system normal tissues (green medians) and in glioma subtypes (red medians). (d) HES6 expression box-plot in glioblastoma samples classified as proneural, neural, classical and mesenchymal glioma subtypes, and as ungrouped samples.

GBMs with strong HES6 expression had amplified P ¼ 0.96 and P ¼ 0.79, respectively). In the Rembrandt PDGFRA (P ¼ 0.005; Supplementary File S2A). EGFR database (Madhavan et al., 2009), HES6 gene expres- amplification or the total EGFR protein level was not sion was not associated significantly with survival in associated with strong HES6 expression (P ¼ 0.693 and GBM (P ¼ 0.290, n ¼ 181) or astrocytoma (P ¼ 0.069, P ¼ 1.0, respectively), but HES6 expression was asso- n ¼ 105). However, in the large TCGA-GBM data set, ciated with expression of phosphorylated EGFR high HES6 gene expression is associated with favorable (P ¼ 0.022; Supplementary File S2B). survival in a Kaplan–Meier analysis with the Anduril tool (P ¼ 0.0354, n ¼ 348; Supplementary File S1D; et al. Tumor HES6 expression and survival in glioblastoma Ovaska , 2010). However, this survival difference can be linked to the different subgroups of GBM. Association of tumor HES6 protein expression with survival was examined among patients diagnosed with either primary or secondary GBM. Survival information Functional importance of HES6 for glioma cell was available from 55 patients. The median time to proliferation death from the date of the diagnosis was 15 months, and HES6 inhibits the cell proliferation when overexpressed all 55 patients eventually died from GBM. Tumor HES6 in HeLa and mouse embryonic fibroblast cells protein expression was not associated with overall (Eun et al., 2008). To evaluate the effects of HES6 in survival when grouped in either three ( þ , þþ or glioma cells, HES6 small interfering (siRNAs) þþþ) or in two categories ( þ and þþ vs þþþ; were transfected into the glioma cell lines A172, LN405,

Oncogene HES6 in glioma S Haapa-Paananen et al 1302 Table 1 HES6 tumor nuclear staining in brain tumor tissues and cell lines. Expression levels of 265 genes were found to histologically normal brain (TMA) be consistently altered when the normalized ratios from Histological type N HES6 expression in both cell lines and from all HES6 siRNA treatments immunohistochemistry were subjected to rank-based meta-analysis (Supple- mentary File S4A). HES6 overexpression was analyzed No Weak Moderate Strong in stably pEYFP-HES6-overexpressing LN405 cells (À)(þ )(þþ)(þþþ) and compared with pEYFP-mock cells. Expression of 372 genes was altered at least twofold after HES6 n (%) n (%) n (%) n (%) overexpression including increase in the progenitor cell Astrocytoma marker nestin (Supplementary File S4B). When both At diagnosis 35 2 (6) 4 (11) 10 (29) 19 (54) HES6 silencing and overexpression data were analyzed Recurrent tumor 15 0 (0) 2 (13) 1 (7) 12 (80) using rank-based meta-analysis, expression of 88 genes Anaplastic astrocytoma was consistently the opposite in knockdown and knock- At diagnosis 31 1 (3) 8 (26) 9 (29) 13 (42) in experiments (Figure 4 and Supplementary File S4C). Recurrent tumor 17 0 (0) 5 (29) 5 (29) 7 (41)a These genes included neurotactin (CX3CL1), transform- Oligodendroglioma ing growth factor-b receptor II (TGFBR2), Norrie At diagnosis 17 0 (0) 7 (41) 5 (29) 5 (29) disease (NDP) and nuclear factor (NF)-kB inhibitor-a Recurrent tumor 15 0 (0) 0 (0) 4 (27) 11 (73) (NFKBIA). Genes that were flagged by HES6 silencing, overexpression or both were analyzed with the Ingenuity Oligoastrocytoma At diagnosis 10 0 (0) 3 (30) 1 (10) 6 (60) Pathways Analysis (IPA, Ingenuity Systems Inc., Recurrent tumor 17 0 (0) 1 (6) 8 (47) 8 (47) Redwood City, CA, USA) and MetaCore (GeneGo Pilocytic astrocytoma 16 0 (0) 3 (19) 3 (19) 10 (63)a Inc., St Joseph, MI, USA) programs. The most Glioblastoma significant biological functions of the commonly de- Primary 116 0 (0) 32 (28) 26 (22) 58 (50) regulated genes included cellular movement, migration, Secondary 20 0 (0) 4 (20) 3 (15) 13 (65) invasion, adhesion, angiogenesis, development, cell-to- Recurrent tumor 30 1 (3) 5 (17) 8 (27) 16 (53) cell signaling, cell proliferation, cell death and apoptosis Total glioma 339 4 (1) 74 (22) 83 (24) 178 (53) (Table 2, and Supplementary File S5).

Non-glioma primary brain tumors Medulloblastoma 18 10 (56) 6 (33) 2 (11) 0 (0) Functional importance of HES6 for glioma cell migration Meningioma 8 5 (63) 2 (25) 1 (13) 0 (0)a As several of the HES6-deregulated genes were related Schwannoma 4 3 (75) 1 (25) 0 (0) 0 (0) to cell migration, a wound-healing assay was conducted Total non-glioma 30 18 (60) 9 (30) 3 (10) 0 (0) Histologically normal 9 6 (66) 3 (33) 0 (0) 0 (0) using the U87MG cell line where HES6 silencing does brain tissue not cause reduction in cell proliferation. HES6 silencing reduced wound closure and wound confluence, indicat- Abbreviations: HES, hairy and enhancer of split 6; TMA, tissue ing inhibition of cell migration (Figure 5). microarray. aThe percentages do not add up to 100 because of rounding. Nuclear staining of HES6 was scored semiquantitatively as:À (no staining), þ HES6 as a transcriptional regulator/cofactor (1–5% positive nuclei), þþ(6–10% positive nuclei) and þþþ(11– The HES6-deregulated genes (listed in Supplementary 100% positive nuclei). File S4) were analyzed using the transcription regulation algorithm in MetaCore (GeneGo Inc.). Aberrant U87MG and into the control cell line SVGp12 expression around key regulator genes, such as myelo- (SV40-transformed fetal astroglial cells). The efficacy cytomatosis viral oncogene homolog (MYC), NF-kB and of knockdown was confirmed by quantitative real-time– p53 was observed (Table 3 and Supplementary File S6). PCR (Figure 3a and Supplementary File S3). Cell HES6 interacts in vitro with a specific type of E-box proliferation, monitored over a 3-day period with the sequence, the ESE box (Jennings et al., 1999; Cossins IncuCyteHD microscope (Essen Instruments Inc., Ann et al., 2002). Significant enrichment of several transcrip- Arbor, MI, USA), was compared with control siRNA- tion factor motifs within the proximal regulatory regions transfected cells. Up to an 80% reduction in the cell of HES6-deregulated genes was apparent, including proliferation rate was seen in A172, 40% in LN405 E-box, MYC, HES1, Sp1 , hepato- and 20% in U87MG and SVGp12 after HES6 cyte nuclear factor 4 (HNF4), p53 and NF-kB siRNA transfection (Figures 3b–d; data not shown). (Supplementary File S7). Interestingly, the HES6 Cell viability, measured by a CellTiter-Glo assay (ATP dimerization partner, HES1 (Bae et al., 2000), binds to measurement, Promega, Madison, WI, USA), was 10 out of the 20 key regulator genes identified (Table 3). reduced by 40–50% in A172 and LN405 cell lines after Furthermore, HES1 expression was reduced after HES6 HES6 silencing (Figure 3e). silencing (Supplementary File S8).

Genome-wide analysis of HES6 silencing and HES6 localizes with promyelocytic leukemia protein overexpression in glioma cell lines and CREB-binding protein The effect of HES6 silencing on genome-wide gene HES6 colocalizes and co-immunoprecipitates with the expression was analyzed in the A172 and LN405 glioma CREB-binding protein (CBP, also known as CREBBP)

Oncogene HES6 in glioma S Haapa-Paananen et al 1303

Figure 2 HES6 immunostaining in non-neoplastic cerebral cortex (a) and diffuse gliomas of various WHO grades. (b) Oligodendroglioma grade 2, HES6 score 1, (c) oligodendroglioma grade 3, score 3, (d) astrocytoma grade 2, score 3, (e) astrocytoma grade 3, score 3 and (f) glioblastoma grade 4, score 3. Arrows point to endothelial cells. Original magnification  400. and the promyelocytic leukemia protein in the nuclear cells, silencing of HES6 did not cause a significant bodies in HeLa cells (Eun et al., 2008). In line with this, change in MYC mRNA levels (data not shown), but a we observed a speckle-like nuclear localization of HES6 in significant reduction of MYC activity took the A172 GBM cell line after transfection with the place when measured using a MYC-luciferase reporter pEYFP-Hes6wt construct. Most of these speckles coloca- assay (P ¼ 0.0039, Figure 6b). Thus, HES6 expression lized with the promyelocytic leukemia-nuclear bodies and can modulate MYC expression and/or activity depend- CBP when co-stained with the respective antibodies ing on the cell line context. (Supplementary File S9). Notably, CBP is known to bind to 17 out of the 20 key regulator genes identified in this study (Table 3). cAMP-response element binding (CREB) motifs were also enriched within the HES6-deregulated Discussion genes (Supplementary File S7). On the basis of genome- and body-wide bioinformatics mining of the GeneSapiens database, HES6 turned out HES6 and MYC to be one of the most glioma-specific genes, showing As the transcription regulation algorithm and the motif overexpression in different histological subtypes enrichment analyses pointed to MYC-responsive genes of glioma as compared with normal brain tissues. This among the HES6-deregulated genes, we also analyzed observation was confirmed in the Rembrandt data MYC expression and promoter activity. In LN405 cells, (Madhavan et al., 2009). High HES6 expression was MYC expression was increased significantly 24 h after found in 23 grade 2–3 astrocytomas and secondary HES6 silencing when assessed using quantitative GBMs following quantitative real-time–PCR analysis real-time–PCR (P ¼ 0.0056, Figure 6a). In U87MG (Somasundaram et al., 2005), but to our knowledge

Oncogene HES6 in glioma S Haapa-Paananen et al 1304 A172 LN405

1.0 A172 LN405 0.9 0.8 0.7 siNEG 0.6 0.5 0.4 0.3

Relative HES6 mRNA 0.2 0.1

siNEG siHES6 siHES6 siHES6 siHES6 12 h 24 h 48 h

100 A172 LN405 100 90 ** 80 * 100 80 ** 90 ** 80 70 60 70 siNEG 60 ** 60 siHES6 50 Death siRNA 40 ** 50 40 *** 40 30 Confluency (%) 30 Cell proliferation (%) Relative confluency (%) 20 *** 20 20 ** 10 10 **

A172 LN405 10 20 30 40 50 60 70 siNEG siHES6 siPLK1 Death siNEG siHES6 Death siRNA Time (hrs) siRNA Figure 3 Silencing of HES6 influences cell proliferation. (a) Quantitative real-time–PCR after HES6 silencing. A172 and LN405 glioblastoma cell lines were transfected with 30 nM siHES6 or negative control siRNA (siNEG), and incubated for 12–48 h. Total RNA was isolated and analyzed with Taqman quantitative real-time–PCR for the expression of HES6 mRNA relative to ACTB. Relative expression is shown with siNEG as a reference. (b) Microscopic images of A172 and LN405 glioblastoma cell lines transfected with 30 nM siHES6 or siNEG, and incubated for 72 h. Images were taken with IncuCyteHD 20x phase optics (Essen Instruments Inc.). (c) Glioma cell lines A172 and LN405 were transfected as above, and cell confluence after 3 days is presented relative to negative control siRNA-transfected cells. (d) Cell proliferation was monitored with IncuCyteHD (Essen Instruments Inc.) time-lapse imaging at 1 h intervals after transfection of the A172 cells with siNEG or siHES6 or cell death siRNA-positive control. Cell confluence was quantified and is shown as monolayer confluence versus time. (e) Silencing of HES6 as above and with an additional positive control siRNA, siPLK1. Cell proliferation was assayed 72 h after transfection using a CellTiter-Glo assay (mean±s.e.m.). Statistically significant changes in panels c and e are shown with asterisks (*Po0.05, **Po0.01, ***Po0.001).

HES6 protein expression has not been studied in clinical with tumor recurrence. HES6 expression is likely not tumors. In our analysis, 98.8% of gliomas showed HES6 useful for making a differential diagnosis among immunoreactivity and 52.5% of gliomas stained oligodendrogliomas, astrocytomas and GBMs, but it strongly, whereas the normal brain tissues stained is a more general biomarker for gliomas. weakly or not at all. In accordance with the mRNA Major molecular features in the proneural subtype expression levels, HES6 protein expression was not include alterations of PDGFRA and isocitrate dehydro- specific to any histological subtype of glioma. However, genase 1 (IDH1) (Verhaak et al., 2010). We found an recurrent grade 2 astrocytomas and grade 2–3 oligoden- association between strong tumor HES6 expression and drogliomas showed higher levels of HES6 immunor- amplifications of PDGFRA, KIT and VEGFR2. HES6 eactivity than the corresponding primary tumors. HES6 interaction partner CBP binds to the promoters of expression was present in endothelial cells in 75% of PDGFRA and VEGFR2 (Illi et al., 2000; Watson et al., glioma samples, suggesting a potential role for this 2002), and we observed that the HES6 overexpression protein in angiogenesis. resulted in increase of PDGFRA and KIT expression Interestingly, high HES6 gene expression was evident (Supplementary File S4). Additionally, 11 genes identi- in proneural tumors and its expression was low in fied in HES6 perturbation experiments are proneural mesenchymal subtypes of GBM. The proneural subtype GBM signature genes (Verhaak et al., 2010), including is associated with more favorable survival as compared neural cell adhesion molecule 1 (NCAM1), SRY-box 11 with the mesenchymal subtype (Cooper et al., 2010; (SOX11), contactin 1 (CNTN1), NK2 2 Verhaak et al., 2010), but the proneural tumors recur (NKX2-2) and microtubule-associated protein 2 often and respond poorly to treatment (Verhaak et al., (MAP2). Strong HES6 expression did not correlate 2010). In line with the association with the proneural with EGFR gene amplification or increased EGFR subtype, high HES6 expression predicted better survival protein expression, both of which are relatively specific in the TCGA-GBM tumors (P ¼ 0.0354) and associated for the classical GBM subtype. Taken together, these

Oncogene HES6 in glioma S Haapa-Paananen et al 1305 50 p53-dependent pathway (Eun et al., 2010). In the

20 present study, transient HES6 overexpression using the 0 pEYFP-HES6wt plasmid resulted in such a high over- –2 02 expression that only glioma cells with modest pEYFP- RPP21 LOC346887 RPL13A HES6wt (approximately 3- to 15-fold) expression DUS3L NOL5A TNFRSF12A survived as a stable cell line. ADM ETS1 Gliomas are invasive brain tumors, and data from the AP4B1 GDF15 FHOD1 HES6 perturbation experiments indicated enrichment in SYTL2 RAB22A CYB5D1 categories related to adhesion, migration CCNE2 HSPC111 MAT2A and invasion. This was supported by the observation X HAS3 RRS1 that HES6 silencing inhibited migration of the U87MG STX1A RND3 cell line (Figure 5). One of the HES6-deregulated SLC20A1 FAM57A NFKBIA genes, previously connected to migration and invasion, TRIB1 BCYRN1 MEG3 was TGFBR2 that suppresses glioma cell migration and LOC338758 NDP DIO2 invasion in vitro and reduces tumorigenicity in vivo after HLA.DRA MOXD1 RNA interference (Naumann et al., 2008; Wesolowska PSG5 GPR177 et al., 2008), and another one was the insulin-like growth PRSS12 CPVL SLC7A3 factor-binding protein 2 (IGFBP2) that increases glioma MOXD1 HLA.DRA CPVL cell migration (Wang et al., 2006). CNTN1 MFAP5 HES6 HES6 has been previously implicated in meta- IGFBP2 CD34 static neuroendocrine prostate cancer, breast cancer OLFML2A COL4A2 RBP1

0 and metastatic colon carcinoma (Swearingen et al., FAM65B GPR177 HPSE 2003; Vias et al., 2008; Hartman et al., 2009). In the SVEP1 HPSE COL11A1 GeneSapiens database, neuroblastoma, melanoma, RPESP COL4A1 SLC2A12 colon cancer and some breast cancers display somewhat EFEMP1 ANXA2P1 elevated expression of HES6, but it is 5–10 times higher MAP2 NFE2L3 EHD4 in the gliomas. HES6 may also have a more general role GFRA1 SUB1 CX3CL1 in cancers. HES6 overexpression increases cell prolifera- PHGDH FBLN1 LAMP2 tion of MCF-7 breast cancer cells in vitro and in TMEM123 SEPT11 NT5C3 xenografts (Hartman et al., 2009). From the genome- SRPX PLD5 wide gene expression profiling after HES6 perturba- C18orf10 C18orf10 TGFBR2 tions, we found aberrant expression around several key WARS CAPNS1 RASGRP3 transcriptional regulator genes. On the basis of chip- CAST IFIT3 SLC44A1 on-chip data (Margolin et al., 2009), HES1 is known to MAP2 KIAA0644 DNASE2 bind to 10 of the 20 key regulator genes found in our LGALS3BP LOC643319 study. Of these key regulator genes, p53 and have HSPA5 AGRN FBLN1 been linked to HES6 previously in other cell types than MFGE8 PGCP FKBP9L glioma (Eun et al., 2008, 2010; Hartman et al., 2009). LGMN CCL26 CTSC The motif enrichment analysis of the HES6-deregulated WARS LAMP2 genes revealed significant enrichment of the same transcriptional regulator motifs, including E-box, MYC, CREB and HES1. HES6 is a transcriptional

siHES6_a siHES6_b siHES6_2 repressor of HES1 (Bae et al., 2000; Gratton et al., 2003), and silencing of HES6 increased HES1 mRNA

pEYFP_HES6wt_b pEYFP_HES6wt_a levels (Supplementary File S8). HES6 colocalizes with Figure 4 Hierarchical clustering of the oppositely deregulated (Supplementary File S9) and also interacts with the genes after HES6 silencing and overexpression. The R/Bioconduc- CREB-binding protein (CREBBP; Eun et al., 2008), tor package RankProd (Hong et al., 2006) was used to find genes which is a transcriptional co-activator and acts as a that are regulated in an opposite way when HES6 is overexpressed (two replicates) or silenced (two replicates of siHES6 and a HES6_2 scaffold to stabilize additional protein interactions siRNA) in LN405 cell line. Hierarchical clustering was performed within the transcription complex. Silencing of HES6 to display the data. and the consequent loss of HES6 from this complex could change expression around the 17 CREBBP- interacting key regulator genes listed in Table 3, findings are compatible with HES6 association with the including MYC. proneural subtype of GBM. E-box and MYC motifs were enriched in the HES6- HES6 likely has an important function in supporting deregulated genes, and silencing of HES6 can affect the growth and survival of glioma cells. In vitro, MYC expression and activity depending on the cell silencing of HES6 inhibited cell proliferation in two context in vitro (Figure 6). On the basis of chip-on-chip GBM cell lines. In primary rat cortical neuron cell data, MYC associates with the HES6 gene promoter lines, overexpression of HES6 using a pEYFP-Hes6wt (P ¼ 0.00182) in a HPB-ALL human cell line (Margolin plasmid causes neuronal death and apoptosis via a et al., 2009), providing a potential feedback loop

Oncogene HES6 in glioma S Haapa-Paananen et al 1306 Table 2 Top 20 common functions from ingenuity pathways analysis after HES6 overexpression and silencing in LN405 Category Function annotation P-value Gene no. Molecules

Cellular movement Movement of 5.22E-07 19 ADM, CAPNS1, CCL26 (includes EG:10344), CD34, COL4A1, normal cells COL4A2, CX3CL1, ETS1, GDF15, GFRA1, HAS3, HPSE, IGFBP2, MAP2, NDP, NFKBIA, TGFBR2, TNFRSF12A, WARS Invasion of cells 2.76E-05 12 ADM, ETS1, FBLN1, GDF15, GFRA1, HAS3, HPSE, IGFBP2, LGMN, MFGE8, RND3, TGFBR2 Cardiovascular Angiogenesis 1.27E-03 8 COL4A1, COL4A2, CX3CL1, ETS1, HPSE, TGFBR2, TNFRSF12A, system development WARS and function Proliferation of en- 1.76E-03 5 ADM, COL4A1, COL4A2, HAS3, NDP dothelial cells Tissue development Adhesion of cells 2.34E-03 12 CD34, CX3CL1, ETS1, GFRA1, HAS3, HPSE, LGALS3BP, MFGE8, NFKBIA, RND3, TGFBR2, TNFRSF12A Cancer Tumorigenesis 1.08E-04 35 AGRN, CAST, CCNE2, CD34, COL4A1, COL4A2, CTSC, CX3CL1, EFEMP1, ETS1, FAM57A, FBLN1, GDF15, GFRA1, HLA-DRA, HPSE, HSPA5, IFIT3, IGFBP2, LGALS3BP, MAT2A, MFAP5, MFGE8, NFKBIA, PHGDH, PRSS12, RBP1, RPL13A, SLC7A3, SRPX, TGFBR2, TNFRSF12A, TRIB1, WARS, WLS Tumor morphology Proliferation of 2.05E-04 3 IGFBP2, NFKBIA, TGFBR2 tumor Cell death Survival of neurons 3.20E-04 6 AGRN, CX3CL1, EHD4, GDF15, GFRA1, NFKBIA Survival of 2.23E-02 10 AGRN, CX3CL1, EHD4, GDF15, GFRA1, HSPA5, NFKBIA, NT5C3, eukaryotic cells RBP1, TGFBR2 Apoptosis of 6.24E-03 19 ADM, CAPNS1, CAST, COL4A2, CX3CL1, DNASE2, ETS1, FBLN1, eukaryotic cells GDF15, GFRA1, HSPA5, LGALS3BP, MFGE8, NFKBIA, RND3, SRPX, STX1A, TGFBR2, TRIB1 Apoptosis 1.29E-02 21 ADM, CAPNS1, CAST, COL4A2, CX3CL1, DNASE2, ETS1, FBLN1, GDF15, GFRA1, HSPA5, LGALS3BP, MFGE8, NFKBIA, RND3, SRPX, STX1A, SUB1, TGFBR2, TNFRSF12A, TRIB1 Cell death 1.21E-02 24 ADM, CAPNS1, CAST, COL4A2, CX3CL1, DNASE2, ETS1, FBLN1, GDF15, GFRA1, HSPA5, IGFBP2, LAMP2, LGALS3BP, MFGE8, NFKBIA, RND3, SRPX, STX1A, SUB1, TGFBR2, TMEM123, TNFRSF12A, TRIB1 Nervous system Survival of neurons 3.20E-04 6 AGRN, CX3CL1, EHD4, GDF15, GFRA1, NFKBIA development and function Cardiovascular dis- Cardiovascular dis- 7.82E-04 27 ADM, CAST, CD34, COL11A1, COL4A1, COL4A2, CPVL, CX3CL1, ease order DIO2, EHD4, ETS1, FAM65B, GDF15, GFRA1, LAMP2, LGMN, MFGE8, MOXD1, NFKBIA, PHGDH, PLD5, RAB22A, SLC44A1, TGFBR2, TNFRSF12A, TRIB1, WARS Organismal develop- Angiogenesis 1.27E-03 8 COL4A1, COL4A2, CX3CL1, ETS1, HPSE, TGFBR2, TNFRSF12A, ment WARS Cellular growth and Proliferation of eu- 6.88E-03 19 ADM, CAPNS1, CCNE2, COL4A1, COL4A2, ETS1, GDF15, HAS3, proliferation karyotic cells HPSE, HSPA5, IGFBP2, MEG3 (includes EG:55384), MFGE8, NDP, NFKBIA, NT5C3, TGFBR2, TNFRSF12A, TRIB1 Proliferation of cells 3.26E-02 20 ADM, CAPNS1, CCNE2, COL4A1, COL4A2, ETS1, GDF15, HAS3, HPSE, HSPA5, IGFBP2, MEG3 (includes EG:55384), MFGE8, NDP, NFKBIA, NT5C3, TGFBR2, TNFRSF12A, TRIB1, WARS Growth of cells 3.76E-02 16 ADM, CAPNS1, CAST, COL4A2, CTSC, CX3CL1, ETS1, GDF15, GFRA1, HAS3, HSPA5, IFIT3, IGFBP2, NFKBIA, TGFBR2, TNFRSF12A Cell-to-cell signaling Binding of 4.81E-03 8 CD34, CX3CL1, GFRA1, HAS3, HPSE, IGFBP2, LGALS3BP, MFGE8 and interaction eukaryotic cells Adhesion of cells 2.34E-03 12 CD34, CX3CL1, ETS1, GFRA1, HAS3, HPSE, LGALS3BP, MFGE8, NFKBIA, RND3, TGFBR2, TNFRSF12A

Abbreviation: HES, hairy and enhancer of split 6.

between HES6 and MYC. Nestin, a marker for 2010). The transcriptional co-activator p300 mediates neuronal stem cells/progenitors and a marker for expression of GFAP and repression of MYC, but MYC poor prognosis in glioma (Strojnik et al., 2007; Schiffer overexpression can over-ride the p300-related effects. et al., 2010), is one of the E-box-containing genes that The p300 silencing increases GBM cell invasion and shows increased expression after HES6 overexpression. results in nestin induction (Panicker et al., 2010). In the MYC overexpression has been reported to activate present study, HES6 silencing resulted in a decrease in transcription of nestin and to repress GFAP, a marker MYC activity and reduced migration of U87MG cells. of differentiated astrocytes, while promoting a more Currently, it is unknown whether HES6 interacts with undifferentiated phenotype in mature murine astrocytes MYC directly or indirectly within a nucleoprotein and GBM cells (Lassman et al., 2004; Panicker et al., complex involving, for example, p300-CBP-CREB.

Oncogene HES6 in glioma S Haapa-Paananen et al 1307 900 org/). From this database (version 3), we systemically sought 800 for putative mRNA expression-based biomarkers in GBM and siNEG anaplastic astrocytoma using a novel Gene Tissue Index 700 siHES6

m) algorithm (Mpindi et al., 2011) to detect outlier gene exp-  600 ression, that is, overexpression of a gene in a subset of the 500 tumor samples. 400 300 Cell culture and reagents

Wound Width ( Cell lines A172 and U87MG were obtained from ECACC 200 (Salisbury, UK), LN405 from the DSMZ (Braunschweig, 100 Germany) and SVG p12 from ATCC (Manassas, VA, USA). Cells were cultured in medium conditions recommended by the 0 5 10 15 20 25 30 providers for less than 4 months before use in these Time after Wounding (h) experiments. Sequences for three siRNAs for HES6 and a control siRNA for PLK1 are described in supplementary 100 materials. All siRNAs were purchased from Qiagen (Hilden, siNEG Germany). Additional control siRNAs were AllStars Hs Cell 90 siHES6 80 Death Control siRNA (Qiagen) and Negative Control siRNA (Qiagen). The siRNAs were used at a final concentration of 70 15 nM each, or as a pool of three siRNAs (denoted as siHES6) 60 10 nM each, thus giving a final concentration of 30 nM. LN405 50 cells were used for creating stable cell lines by transfecting 40 pEYFP-C1-mock and pEYFP-C1-HES6 with Fugene6 (Roche 30 Applied Science, Indianapolis, IN, USA) and selecting the

Wound Confluency % 20 positive cells with 600 mg/ml and maintained with 400 mg/ml 10 of G418 (Sigma-Aldrich, St Louis, MO, USA).

0 5 10 15 20 25 30 Real-time quantitative PCR analysis Time after wounding (h) Total cellular RNA was isolated using a MiRVana Total RNA Figure 5 Wound assay in U87MG cells. At 24 h after HES6 was isolation kit (Ambion, Austin, TX, USA) or an RNeasy Mini silenced in U87MG cell line, scratch wounds were made, and cells kit (Qiagen). For complementary DNA synthesis, 200 ng of were monitored by taking time-lapse-phase-contrast images (three RNA was reverse transcribed with High Capacity cDNA images/well) once every hour with an IncuCyteHD instrument. Reverse Transcription kit (Applied Biosystems, Foster City, Wound width (mm) and wound confluence (%) are shown in the CA, USA). The complementary DNA was diluted 1:2 and the two panels. The assay was repeated twice with triplicates. Taqman quantitative real-time–PCR analysis was performed with an Applied Biosystems 7900HT instrument using specific primers for HES6, HES1, MYC, glyceraldehyde 3-phosphate dehydrogenase and actin beta (ACTB) designed with the In the presence of gliogenic stimuli, HES6 over- Universal Probe Library Assay Design Center (Roche Applied expression increases nestin and decreases GFAP levels Science). The sequences of the primers are described in in cortical progenitor cells. HES6 expression prevents supplementary materials. The fluorescent Taqman probes the precursor-to-astrocyte transition, resulting in an were obtained from Roche Human Probe Library. The results expansion of the neural progenitor cell pool (Jhas et al., were analyzed with SDS 2.3 and RQ manager software (Applied Biosystems), and expression of HES6, MYC or 2006). In glioma, HES6 overexpression might maintain HES1 mRNA was determined by relative quantification the cancer population as suggested by the method using ACTB or glyceraldehyde 3-phosphate dehydro- increase of nestin expression, although HES6 and nestin genase as an endogenous control. The data were collected from or CD133 protein levels were not associated in the 2–3 separate biological experiments, which were run twice in current study. quadruplicates. We conclude that HES6 is a key molecular modulator in glioma that likely influences tumor growth and cell Immunohistochemistry migration. HES6 is selectively and frequently over- Formalin-fixed, paraffin-embedded tissues from 417 brain expressed in glioma and is an important transcriptional tumors of patients who underwent craniotomy for a primary regulator. or recurrent glioma in 1979–2003 were retrieved from the archives of the Department of Pathology, Helsinki University Central Hospital, Finland, as described previously (Joensuu et al., 2005; Puputti et al., 2006) and from the Department of Materials and methods Pathology, Tampere University Hospital (Haapasalo et al., 2006), for construction of a TMA (Table 1). Whole-tissue Data mining from the GeneSapiens database sections of histologically normal brain tissue (n ¼ 9) were The GeneSapiens database contains gene expression data of analyzed. For detailed description of the TMA construction, 17 330 human genes in 9783 samples from 175 types of healthy see supplementary materials and methods. and pathological tissues (Kilpinen et al., 2008). This database Five-micrometer sections were cut for HES6 immunohis- is comprised of Affymetrix gene expression array experiments tochemistry. HES6 expression was analyzed on tumor TMA, and allows evaluation of mRNA expression levels of most and non-tumor and normal whole-tissue sections. HES6 was human genes across different tissues (http://www.genesapiens. detected with a synthetic peptide derived from the C terminus

Oncogene HES6 in glioma S Haapa-Paananen et al 1308 Table 3 Aberrant expression around transcriptional key regulator genes after HES6 silencing, overexpression or both when analyzed using the transcription regulation algorithm in MetaCore Key regulator genes HES6, all LN405 HES6 LN405 HES6 siRNA HES1-binding CREBBP /CBP siRNAs overexpression and overexpression chip dataa interactionb

Altered P-value Altered P-value Altered P-value genes genes genes

SP1 57 4.74E-135 131 7.58E-298 — — Yes — c-Myc 53 2.51E-125 45 4.39E-99 15 1.85E-42 Yes Yes HNF4-a 49 3.19E-113 41 3.98E-90 14 1.38E-39 Yes Yes p53 35 5.21E-82 64 5.71E-142 13 1.02E-36 Yes Yes NF-kB 31 1.76E-72 58 2.29E-128 17 3.12E-48 Yes — STAT1 25 2.93E-58 33 2.73E-72 — — — Yes ESR1 (nuclear) 24 6.75E-56 66 1.62E-146 14 1.38E-39 — Yes c-Jun 19 3.98E-44 46 2.51E-101 — — Yes RelA (p65 NF-kB subunit) 19 3.98E-44 37 3.40E-81 6 6.59E-17 — Yes C/EBP-b 17 1.95E-39 34 4.69E-76 6 6.59E-17 — Yes SRF 16 4.25E-37 — 6 6.59E-17 — Yes 15 9.24E-35 45 4.39E-99 — — — Yes EGR1 12 8.98E-28 34 1.63E-74 8 1.62E-22 Yes Yes GCR-a (NR3C1) 10 3.90E-23 38 2.00E-83 5 4.01E-14 Yes Yes GATA-1 — — 42 2.30E-92 10 3.60E-28 — Yes VDR — — 31 7.58E-68 4 2.34E-11 — Yes E2F1 15 9.24E-35 — — — — Yes Yes c-Fos — — — — 5 4.01E-14 Yes Yes SP3 10 3.90E-23 — — — — Yes — CREB1 — — 55 1.38E-121 — — — Yes

Abbreviations: CREBBP, CREB-binding protein; HES, hairy and enhancer of split; siRNA, small interfering RNA. Data sources for the last two columns: aMargolin et al., 2009. bProtein Interaction Network Analysis (Wu et al., 2009; http://csbi.ltdk.helsinki.fi/pina/).

of Human HES6 rabbit polyclonal antibody (ab66461, confluence was monitored by taking time-lapse phase-contrast Abcam, Cambridge, UK; dilution 1:1000, incubation o/n images (3 images/well) once per hour of live cells grown for þ 4 1C). Staining was done with a PowerVision staining kit 30 hr in a CO2 incubator with an IncuCyteHD microscope (ImmunoVision Technologies Co., Daly City, CA, USA) (Essen Instruments Inc.). (Sihto et al., 2009). The tissue sections were counter stained with hematoxylin. Illumina gene expression and data analyses HES6 immunoreactivity in TMA samples was evaluated by A172 and LN405 cells (240 000/well on six-well plates) were two pathologists (HH and OT) using a consultation micro- transfected for 12 h with a siHES6 pool, three individual scope (Nikon Eclipse E600, Nikon Instech Co., Kanagawa, siRNAs or negative control siRNA at 30 nM using SiLentFect Japan). Nuclear staining of HES6 in tumor cells in whole- (Bio-Rad Laboratories, Hercules, CA, USA). One to two tissue core area (0.28 mm2 or 0.79 mm2) was scored semiquan- biological replicate transfections were performed. Total RNA titatively by consensus of the two investigators as: 0 (no was isolated with an RNeasy kit (Qiagen). RNA quality was staining), 1 (1–5% positive nuclei), 2 (6–10% positive nuclei) evaluated using an Agilent 2100 Bioanalyzer (Agilent Tech- or 3 (11–100% positive nuclei). Endothelial cell immunor- nologies, Palo Alto, CA, USA). In all, 300 ng of purified total eactivity was classified as either negative or positive (staining RNA was amplified with an Illumina Total Prep RNA in one or more nuclei). Amplification kit (Illumina, San Diego, CA, USA), and 750 ng of biotin-labeled complementary RNA was hybridized Proliferation assays to Sentrix HumanHT-12 v3 Expression BeadChips (Illumina) Cell proliferation was assayed with a CellTiter-Glo Cell at the Finnish DNA Microarray Centre, Turku Centre for Viability assay (Promega, Madison, WI, USA). The signals Biotechnology. The arrays were scanned with an Illumina were quantified using an Envision Multilabel Plate Reader BeadArray Reader, and raw data were produced using a (Perkin-Elmer, Massachusetts, MA, USA). Cell growth and GenomeStudio (Illumina). The microarray data have been confluence was monitored by taking time-lapse phase-contrast deposited in the NCBI’s Gene Expression Omnibus and is images once per hour from cells grown for 3 days in a CO2 accessible through a GEO Series accession number GSE22692. incubator using an IncuCyteHD microscope (Essen Instru- The Illumina data analyses are described in detail in the ments Inc.). supplementary methods.

Wound assay MYC-luciferase reporter assay U87MG cells (150 000 cells/well) were reverse transfected with The activity of Myc-regulated signals transduction pathways 30 nM siHES6 or negative control siRNA using SiLentFect in were assessed using a Cignal MYC reporter assay kit 24-well Image Lock plates (Essen Instruments Inc.). Wounds (SABiosciences, Qiagen) according to the manual. In brief, were generated next day by scratching with a 24-well Wound U87MG cells (20 000/well) were co-transfected either with maker tool (Essen Instruments Inc.). Wound closure and 100 ng of (1) an inducible MYC-responsive firefly luciferase

Oncogene HES6 in glioma S Haapa-Paananen et al 1309 1.8 construct, (2) a non-inducible negative construct, (3) positive HES6 ** control (a mixture of constitutively expressing firefly luciferase, 1.6 MYC constitutively expressing renilla luciferase and constitutively expressing GFP 1:1:40), together with 30 nM negative control 1.4 siRNA or siHES6 using 0.4 mL of Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) in a white, clear-bottom 1.2 96-well plates. Firefly and renilla luciferase activities were assayed 48 h after transfection with a Dual-Glo Luciferase 1.0 Assay System (Promega) using an Envision Plate Reader (Perkin-Elmer). 0.8 Statistical analysis 2 0.6 Frequency tables were analyzed with the w -test or Fisher’s exact test. Survival was analyzed using the Kaplan–Meier method, and 0.4 *** survival of groups was compared using the log-rank test. Overall survival was calculated from the date of the diagnosis to the date Relative mRNA level in LN405 cells *** 0.2 of death, and the patients alive were censored on the last date of last follow-up. All P-values are two-sided. 0.0 siNEG siHES6_24h siHES6_48h Conflict of interest 25 siNEG The authors declare no conflict of interest. siHES6

Acknowledgements 20 This work was supported by the Academy of Finland Translational Genome-Scale Biology Center of Excellence, the Sigrid Juselius foundation and EU-FP6 project RIGHT 15 (LSHB-CT-2004-005276). We acknowledge Dr Bokkee Eun for the generous gift of the HES6 overexpression plasmid pEYFP-C1-Hes6 (Eun et al., 2008). We also acknowledge the ** Finnish DNA Microarray Centre for their excellent technical assistance. 10

References Relative luciferase activity

Bae S, Bessho Y, Hojo M, Kageyama R. (2000). The bHLH gene 5 Hes6, an inhibitor of Hes1, promotes neuronal differentiation. Development 127: 2933–2943. Cancer Genome Atlas Research Network (2008). Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature 455: 1061–1068. 0 Cooper LA, Gutman DA, Long Q, Johnson BA, Cholleti SR, Kurc T Neg control MYC reporter et al. (2010). The proneural molecular signature is enriched in Figure 6 MYC promoter activity and expression in HES6- oligodendrogliomas and predicts improved survival among diffuse silenced cells. (a) Quantitative real-time–PCR of MYC and HES6 gliomas. PLoS One 5: e12548. after HES6 silencing. LN405 glioblastoma cell line was transfected Cossins J, Vernon AE, Zhang Y, Philpott A, Jones PH. (2002). Hes6 with 30 nM siHES6 or negative control siRNA (siNEG), regulates myogenic differentiation. Development 129: 2195–2207. and incubated for 24 or 48 h. Thereafter, the total cellular RNA Eun B, Cho B, Moon Y, Kim SY, Kim K, Kim H et al. (2010). was isolated and analyzed with Taqman quantitative real-time– Induction of neuronal apoptosis by expression of Hes6 via PCR for the expression of MYC and HES6 mRNA. The results p53-dependent pathway. Brain Res 1313: 1–8. normalized for ACTB are shown as relative expression using Eun B, Lee Y, Hong S, Kim J, Lee HW, Kim K et al. (2008). Hes6 siNEG as a reference. Statistically significant changes are shown controls cell proliferation via interaction with cAMP-response with asterisks (**Po0.01, ***Po0.001, n ¼ 5 biological repeats). element-binding protein-binding protein in the promyelocytic (b) A firefly luciferase MYC reporter assay was used to measure the leukemia nuclear body. J Biol Chem 283: 5939–5949. MYC transcription promoter activity when HES6 was silenced. Gratton MO, Torban E, Jasmin SB, Theriault FM, German MS, The U87MG cells were co-transfected using Lipofectamine2000 Stifani S. (2003). Hes6 promotes cortical and inhibits with MYC reporter or negative control, together with 30 nM Hes1 transcription repression activity by multiple mechanisms. siHES6 or siNEG. Luciferase activity was measured after 48 h Mol Cell Biol 23: 6922–6935. incubation with Dual-Glo Luciferase assay. The MYC promoter Haapasalo JA, Nordfors KM, Hilvo M, Rantala IJ, Soini Y, activity is expressed as arbitrary units using a renilla reporter Parkkila AK et al. (2006). Expression of carbonic anhydrase IX for internal normalization. Results are averages of three in astrocytic tumors predicts poor prognosis. Clin Cancer Res 12: separate experiments, each conducted in triplicate ( þ /Às.e.m., 473–477. **Po0.01).

Oncogene HES6 in glioma S Haapa-Paananen et al 1310 Hartman J, Lam EW, Gustafsson JA, Strom A. (2009). Hes-6, an Ohgaki H, Kleihues P. (2009). Genetic alterations and signaling inhibitor of hes-1, is regulated by 17beta-estradiol and promotes pathways in the evolution of gliomas. Cancer Sci 100: 2235–2241. breast cancer cell proliferation. Breast Cancer Res 11: R79. Ohgaki H, Kleihues P. (2007). Genetic pathways to primary and Hong F, Breitling R, McEntee CW, Wittner BS, Nemhauser JL, secondary glioblastoma. Am J Pathol 170: 1445–1453. Chory J. (2006). RankProd: A bioconductor package for detecting Ovaska K, Laakso M, Haapa-Paananen S, Louhimo R, Chen P, differentially expressed genes in meta-analysis. Bioinformatics 22: Aittomaki V et al. (2010). Large-scale data integration framework 2825–2827. provides a comprehensive view on glioblastoma multiforme. Illi B, Puri P, Morgante L, Capogrossi MC, Gaetano C. (2000). Genome Med 2: 65. Nuclear factor-kappaB and cAMP response element binding Panicker SP, Raychaudhuri B, Sharma P, Tipps R, Mazumdar T, protein mediate opposite transcriptional effects on the flk-1/KDR Mal AK et al. (2010). p300- and myc-mediated regulation of gene promoter. Circ Res 86: E110–E117. glioblastoma multiforme cell differentiation. Oncotarget 1: Jennings BH, Tyler DM, Bray SJ. (1999). Target specificities 289–303. of drosophila enhancer of split basic helix-loop-helix . Puputti M, Tynninen O, Sihto H, Blom T, Maenpaa H, Isola J et al. Mol Cell Biol 19: 4600–4610. (2006). Amplification of KIT, PDGFRA, VEGFR2, and EGFR in Jhas S, Ciura S, Belanger-Jasmin S, Dong Z, Llamosas E, gliomas. Mol Cancer Res 4: 927–934. Theriault FM et al. (2006). Hes6 inhibits astrocyte differentiation Schiffer D, Annovazzi L, Caldera V, Mellai M. (2010). On the origin and promotes neurogenesis through different mechanisms. and growth of gliomas. Anticancer Res 30: 1977–1998. J Neurosci 26: 11061–11071. Sihto H, Tynninen O, Halonen M, Puputti M, Karjalainen-Lindsberg Joensuu H, Puputti M, Sihto H, Tynninen O, Nupponen NN. (2005). ML, Kukko H et al. (2009). Tumour microvessel endothelial Amplification of genes encoding KIT, PDGFRalpha and VEGFR2 cell KIT and stem cell factor expression in human solid tumours. receptor tyrosine kinases is frequent in glioblastoma multiforme. Histopathology 55: 544–553. J Pathol 207: 224–231. Somasundaram K, Reddy SP, Vinnakota K, Britto R, Subbarayan M, Kilpinen S, Autio R, Ojala K, Iljin K, Bucher E, Sara H et al. (2008). Nambiar S et al. (2005). Upregulation of ASCL1 and inhibition of Systematic bioinformatic analysis of expression levels of 17 330 characterize progressive astrocytoma. human genes across 9 783 samples from 175 types of healthy and Oncogene 24: 7073–7083. pathological tissues. Genome Biol 9: R139. Strojnik T, Rosland GV, Sakariassen PO, Kavalar R, Lah T. (2007). Koyano-Nakagawa N, Kim J, Anderson D, Kintner C. (2000). Hes6 markers, nestin and musashi proteins, in the acts in a positive feedback loop with the to promote progression of human glioma: Correlation of nestin with prognosis neuronal differentiation. Development 127: 4203–4216. of patient survival. Surg Neurol 68: 133–143. Kumar HR, Zhong X, Sandoval JA, Hickey RJ, Malkas LH. (2008). Swearingen ML, Sun D, Bourner M, Weinstein EJ. (2003). Detection Applications of emerging molecular technologies in glioblastoma of differentially expressed HES-6 gene in metastatic colon carcino- multiforme. Expert Rev Neurother 8: 1497–1506. ma by combination of suppression subtractive hybridization and Lassman AB, Dai C, Fuller GN, Vickers AJ, Holland EC. (2004). cDNA library array. Cancer Lett 198: 229–239. Overexpression of c-MYC promotes an undifferentiated phenotype Verhaak RG, Hoadley KA, Purdom E, Wang V, Qi Y, Wilkerson MD in cultured astrocytes and allows elevated ras and akt signaling et al. (2010). Integrated genomic analysis identifies clinically to induce gliomas from GFAP-expressing cells in mice. Neuron relevant subtypes of glioblastoma characterized by abnormalities Biol 1: 157–163. in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 17: 98–110. Madhavan S, Zenklusen JC, Kotliarov Y, Sahni H, Fine HA, Buetow Vias M, Massie CE, East P, Scott H, Warren A, Zhou Z et al. (2008). K. (2009). Rembrandt: Helping personalized medicine become a Pro-neural transcription factors as cancer markers. BMC Med reality through integrative translational research. Mol Cancer Res 7: Genomics 1: 17. 157–167. Wang GK, Hu L, Fuller GN, Zhang W. (2006). An interaction Margolin AA, Palomero T, Sumazin P, Califano A, Ferrando AA, between insulin-like growth factor-binding protein 2 (IGFBP2) Stolovitzky G. (2009). ChIP-on-chip significance analysis reveals and integrin alpha5 is essential for IGFBP2-induced cell mobility. large-scale binding and regulation by human transcription factor J Biol Chem 281: 14085–14091. oncogenes. Proc Natl Acad Sci U S A 106: 244–249. Watson PA, Vinson C, Nesterova A, Reusch JE. (2002). Content Mpindi JP, Sara H, Haapa-Paananen S, Kilpinen S, Pisto T, Bucher E and activity of cAMP response element-binding protein regulate et al. (2011). GTI: A novel algorithm for identifying outlier platelet-derived growth factor receptor-alpha content in vascular gene expression profiles from integrated microarray datasets. smooth muscles. Endocrinology 143: 2922–2929. PLoS One 6: e17259. Wesolowska A, Kwiatkowska A, Slomnicki L, Dembinski M, Master Naumann U, Maass P, Gleske AK, Aulwurm S, Weller M, Eisele G. A, Sliwa M et al. (2008). Microglia-derived TGF-beta as an (2008). Glioma gene therapy with soluble transforming growth important regulator of glioblastoma invasion–an inhibition of factor-beta receptors II and III. Int J Oncol 33: 759–765. TGF-beta-dependent effects by shRNA against human TGF-beta Noushmehr H, Weisenberger DJ, Diefes K, Phillips HS, Pujara K, type II receptor. Oncogene 27: 918–930. Berman BP et al. (2010). Identification of a CpG island methylator Wu J, Vallenius T, Ovaska K, Westermarck J, Makela TP, Hautaniemi phenotype that defines a distinct subgroup of glioma. Cancer Cell S. (2009). Integrated network analysis platform for protein-protein 17: 510–522. interactions. Nat Methods 6: 75–77.

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

Oncogene