Identification of Differentially Expressed and Developmentally

Identiﬁcation of differentially expressed and developmentally regulated genes in medulloblastoma using suppression subtraction hybridization

Naoki Yokota1,2,3, Todd G Mainprize1,2, Michael D Taylor1,2, Tomohiko Kohata1,2, Michael Loreto1,2, Shigeo Ueda1,2, Wieslaw Dura4, Wiesia Grajkowska4, John S Kuo1,2 and James T Rutka1,2,*

1The Arthur and Sonia Labatt Brain Tumor Research Centre, The University of Toronto, Toronto, Ontario, Canada; 2The Division of Neurosurgery, The University of Toronto, Toronto, Ontario, Canada; 3The Department of Neurosurgery, Hamamatsu University School of Medicine, Hamamatsu, Japan; 4Department of Pathology, University of Warsaw, Warsaw, Poland

To increase our understanding of the molecular pathogen- Keywords: medulloblastoma; developmentally regulated esis of medulloblastoma (MB), we utilized the technique of gene; differentially expressed gene; subtractive suppres- suppression subtractive hybridization (SSH) to identify sion hybridization; cerebellum; external granule cell genes that are dysregulated in MB when compared to cerebellum. SSH-enriched cDNA libraries from both human and Ptch þ /À heterozygous murine MBs were generated by subtracting common cDNAs from corre- Introduction sponding non-neoplastic cerebellum. For the human classic MB library, total human cerebellar RNA was Despite the paucity of knowledge regarding the mole- þ /À used as control tissue; for the Ptch heterozygous MB, cular oncogenesis of most human brain tumors, our þ /À non-neoplastic cerebellum from an unaffected Ptch understanding of the molecular pathways causing littermate was used as the control. Through differential medulloblastoma (MB), the most common malignant screening of these libraries, over 100 upregulated tumor brain tumor in children, has advanced considerably. We cDNA fragments were isolated, sequenced and identified now know that MB can occur in the setting of two with the NCBI BLAST program. From these, we selected major inherited genetic syndromes. In these two genes involved in cellular proliferation, antiapoptosis, and hereditary diseases, Turcot and Gorlin syndrome, the cerebellar differentiation for further analysis. Upregulated Wnt and Hedgehog (Hh) signaling pathways respec- genes identified in the human MB library included Unc33- tively, are dysregulated and may lead to MB as well as like protein (ULIP), SOX4, Neuronatin (NNAT), the other developmental malformations and cancers (Taylor mammalian homologue of Drosophila BarH-like et al., 2000; Wechsler-Reya and Scott, 2001; Yokota 1(BARHL1), the nuclear matix protein NRP/B et al., 2002). MB is also found in approximately 20% of (ENC1), and the homeobox OTX2 gene. Genes found to offspring from heterozygous transgenic mice carrying a be upregulated in the murine MB library included cyclin targeted deletion of the Ptch gene that results in D2 (Ccnd2), thymopoietin (Tmpo), Musashi-1 (Msh1), constitutive Hh signaling (Goodrich et al., 1997; protein phosphatase 2A inhibitor-2 (I-2pp2a), and Wetmore et al., 2001). Both Wnt and Hh signaling Unc5h4(D). Using semiquantitative reverse transcrip- pathways play crucial roles in the proliferation and tion–polymerase chain reaction (RT–PCR), the mRNA differentiation of cerebellar granule cells, from which expression levels for these genes were markedly higher in MBs are thought to arise (Yokota et al., 1996; Buhren human MBs than in cerebellum. Western blot analysis was et al., 2000). used to further confirm the overexpression of a subset of Primitive neuroectodermal tumors (PNETs), such as these genes at the protein level. Notch pathway over- MB, are characterized by dysregulation of cellular activity was demonstrated in the TE671 MB cell line proliferation, apoptosis, or differentiation signals that expressing high levels of MSH1 through HES1-Lucifer- normally occur at critical developmental stages in ase transfections. This study has revealed a panel of neurogenesis (Wechsler-Reya and Scott, 2001). By developmentally regulated genes that may be involved in histology, MB is composed mainly of undifferentiated the pathogenesis of MB. small, round blue cells. However, many MBs exhibit Oncogene (2004) 23, 3444–3453. doi:10.1038/sj.onc.1207475 markers of neuronal differentiation including expression Published online 5 April 2004 of neuronal transcriptional factors from the PAX (Kozmik et al., 1995; Vincent et al., 1996), ZIC (Yokota *Correspondence: JT Rutka, The Division of Neurosurgery, Suite et al., 1996; Michiels et al., 1999; Pomeroy et al., 2002), 1502, The Hospital for Sick Children, 555 University Avenue, and NEUROD gene families (Rostomily et al., 1997). Toronto, Ontario, Canada M5G 1X8; E-mail: [email protected] These neuronal transcriptional factors are developmen- Received 7 September 2003; revised 5 January 2004; accepted 5 January tally regulated and govern the expression of several 2004; Published online 5 April 2004 genes that influence cell fate in the cerebellar granule cell Developmentally regulated genes in medulloblastoma N Yokota et al 3445 lineage, and presumably affect the growth and differ- the fragments isolated from the human MB cDNA entiation of MB. In this study, we sought to identify library, most were found to bear homology to known genes that are differentially expressed in a sporadic sequences in the human genome database. Other cDNA human classical MB as compared with normal human fragments could be assigned a genomic location and a cerebellar tissue; and in a murine Ptch þ /À MB as hypothetical protein structure. However, the structure compared with non-neoplastic cerebellar tissue from and function of some human ESTs found through an unaffected littermate. We utilized the technique of homology searches are still largely unknown. Among suppressive subtractive hybridization (SSH), a polymer- the 25 fragments isolated from the murine cDNA ase chain reaction (PCR)-based method of subtractive subtractive library, 17 identified genes were repeat hits cloning, to identify genes overexpressed in the two with Ccnd2 being represented 12 times (almost 50% of different types of MB (Diatchenko et al., 1996). all clones). This may be due to a high native number of Although a number of genes were isolated using this mRNA species, or related to the preamplification step strategy, we concentrated our efforts on those genes that used for generating the murine library. We selected five are known to be developmentally regulated in the of the nine different genes identified in the murine central nervous system (CNS). Confirmation of the cDNA library for verification. results obtained from the SSH screen was performed by In both libraries, we identified many developmentally reverse transcription (RT)–PCR and Western blot regulated genes involved in cellular proliferation, anti- analysis of non-neoplastic cerebellum, MB specimens apoptosis, and differentiation of the cerebellar granule and various cell lines. Our results show that many MBs cells (Table 2a, b). Housekeeping, ribosomal and share certain expression patterns of these developmen- mitochondrial genes were rarely isolated, indicating that tally regulated genes. the subtractive hybridizations were effective. A total of 11 clones of differentially expressed and developmentally regulated genes were selected from both MB libraries and subjected to further analysis by Results RT–PCR (Figure 1). The genes from the human MB Isolation of upregulated clones by screening subtracted library encode for the nuclear matrix protein NRP/B cDNA libraries (ENC1), the homeobox protein OTX2, SRY (sex determination region Y)-box 4 protein SOX4, Neuro- Over 300 clones were screened in each library, from natin (NNAT), human homologue of Drosophila which 57 differentially regulated clones were selected. BARHL1, and Unc-33-like protein (ULIP). The genes Their plasmids were then purified and cDNAs se- selected from the murine cDNA library are cyclin D2 quenced. The percentage of upregulated clones was (CCND2), thymopoietin (TMPO), Musashi-1 therefore approximately 20%. Of the 32 clones identified (MSH1), protein phosphatase 2A inhibitor-2 from the human MB library, 12 were multiple hits and (I-2PP2A), which is a murine homolog of the human we selected six for further evaluation (Table 1). Among SET gene, and the candidate netrin receptor UNC5H4.

Table 1 Primer sequences of genes of interest Primer name Sequence Ann. temperature (1C) Length (bp)

BARHL1 forward GGAAGGGACTGTTTGGAGAC 58 318 BARHL1 reverse GCCCTCCTCCTTCACTTTAT ULIP forward GGCAGAAGCAAGAAGAGATT 49 196 ULIP reverse CAAACAAATCAAGGCTATGC SOX4 forward ATTGATGTTGTTGTTGATGG 53 196 SOX4 reverse AAGCAAAATAAAACAAAACC OTX2 forward AGTCACCAGCCATCTCAATC 51 156 OTX2 reverse ATAATCCAAGCAGTCAGCAT Neuronatin forward CGACAATGACGAAGATACCA 53 131 Neuronatin reverse ATCAGAATGCGGTGCCTATG ENC1 forward TGGCCATGGAGGAACTCATC 57 635 ENC1 reverse TGGGGAGCTTGTCATGACTG CCND2 forward ATTGGCATGTCTGGTTCACA 58 195 CCND2 reverse GAACGCCAGATACCAGAAGC TMPO forward CAGGTTCCTTTGTGGCATTT 58 202 TMPO reverse TTGCTGCCATTCTTCTTCAA MSH-1 forward ATGGTGGAATGTAAGAAAGC 58 205 MSH-1 reverse TCGGGGAACTGGTAGGTGTA SET forward CAGCAAGAAGCGATTGAACA 58 203 SET reverse TGCAGACACTTGTGGATGGT Unc5H4 forward GGTGCTCCTGAGTCCTGAAG 58 203 Unc5H4 reverse GGGTCCAAAAGGCAGTAACA b-Actin forward CAACCGCGAGAAGATGACC 58 326 b-Actin reverse TCCAGGGCGACATAGCACA

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3446 Table 2 (a) Gene identiﬁcation from human MB library. (b) Identiﬁed genes from murine MB library Gene name Accession no. Hit no.

(a) Dihydropyrimidinase-related protein-3 (Unc33 like protein; ULIP) D78014 4 Okadaic acid-inducible phosphoprotein (OA48-18) AF069250 3 Cisplatin resistance-associated overexpressed protein AB034205 GAP-associated tyrosine phosphoprotein p62 NM_006559 3 Synaptotagmin IV (SYT4) XM_031161 3 Nonhistone chromosomal protein HMG-14 J02621 2 SRY-box containing gene 4 (Sox4) AF124147 2 Transthyretin (prealbumin, amyloidosis type I) (TTR) BC005310 2 Neuronatin a and b U31767 1 Nuclear matrix protein NRP/B (NRPB)/ (ENC1) AF059611 1 Homeobox protein OTX2 AF298117 1 BarH (Drosophila)-like 1 (BARHL1) NM_020064 1 Membrane glycoprotein M6 (GPM6a) D49958 1 Vitiligo-associated protein VIT-1 AF264714 1 Transcription factor ETR101 M62831 1 Isoform of CLC-6 chloride channel X96391 1 Substance P beta-PPT-A (TAC1) NM_013998 1 Microﬁbril-associated glycoprotein (MFAP2) U19718 1 50 cap guanine-N-7 methyltransferase AF067791 1 Spliceosomal protein SAP 155 AF054284 1 PR264 gene splicing factor SFRS2 NM_003016 1

(b) G1/S-speciﬁc cyclin D2 NM_009829 12 Tubulin a-1 chain NM_011653 4 Thymopoietin beta U39074.1 3 Cytoskeletal g-actin M21495.1 2 Retrovirus-like intracistronic type A particle element M18252.1 1 Musashi-1 gene for RNA-binding protein D49654.1 1 Germline immunoglobulin g-1 (Ighb) L35252.1 1 Protein phosphatase 2A inhibitor-2, I-2PR2A AB044937 1 Transmembrane receptor Unc5H4 NM_022207 1

Figure 1 RT–PCR of genes of interest in human MB specimens. RT–PCR analysis of the human MB tumor samples revealed three different expression patterns for the 11 genes examined: (1) genes that were largely coexpressed in the same pattern across all the tumor samples examined referable to non-neoplastic cerebellum (BARHL1, NNAT, SOX4, UNC4H5, CCND2); (2) genes that were overexpressed throughout all the tumor samples (ULIP, ENC1, SET, TMPO, MSH1); and (3) a gene that had no speciﬁc pattern of expression in the tumor samples (OTX2) b-Actin demonstates equal expression across the panel. MB1–25: medulloblastoma case 1–25, hCB- non-neoplastic human cerebellum

RT–PCR analysis of MB cell lines and specimens samples than in normal cerebellum by semiquantitative RT–PCR. In most MB samples examined, such over- To be considered upregulated in MBs, the tested genes expression was found (Figure 1). Moreover, there had to show a threefold greater expression in tumor

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3447 seemed to be some patterns of expression between the the MB tumor specimens. In the second group, TMPO different MB cases. For example, in case MB2 and and SET were also expressed throughout the panel of MB7, in which there was high expression of GLI1 MB tumors and could represent genes that are transcripts indicative of Hh activation (data not shown), important in a proliferative state regardless of tumor OTX2, SOX4, NNAT, and ENC1 showed decreased phenotype. expression. Interestingly, these were all identified in the human library and not in the Ptch þ /À heterozygous Analysis of notch and WNT pathways in MB specimens mouse MB library. and cell lines RT–PCR analysis of the human MB tumor samples revealed three different expression patterns for the 11 CCND2 and MSH1 were analysed by Western blot as genes examined: (1) genes that were largely coexpressed they were found to have elevated transcripts by RT– in a similar pattern (upregulated or not expressed) PCR (Figure 3). CCND2 expression was elevated in the across all the tumor samples examined referable to non- PFSK and TE671 cell lines and in three of the six MB neoplastic cerebellum (BARHL1, NNAT, SOX4, specimens, relative to non-neoplastic cerebellum. While UNC4H5, CCND2); (2) genes that were overexpressed it is known that CCND2 is upregulated via the HH throughout all the tumor samples ULIP, ENC1, SET, pathway in various MB tumors and cell lines, we also TMPO, MSH1; and (3) a gene that had no specific wanted to analyse the expression of CCND1, a WNT- pattern of expression in the tumor samples (OTX2) regulated cyclin D, in the same samples. CCND1 was (Figure 1). The lack of expression of all of these genes in moderately expressed in all but one MB specimens and the non-neoplastic cerebellum is further evidence of the in six of eight MB cell lines. As CCND1 is known to be a efficacy of the SSH technique. The uniform expression direct target of CTNNB1 (b-catenin), we were not of the b-actin control gene indicates good integrity and surprised by their correlative expression in the MB cell equal template loading of the RNA from the tumor lines. Interestingly, there was little correlation in specimens. expression of CCND1 and b-catenin in the MB tumor The expression profile of the 11 genes in the array of samples. human tumor cell lines were also categorized into three MSH1 protein was not found in cerebellum, but was groups: (1) genes that were not primarily expressed by observed strongly in TE671 and D341 and to a lesser any of the cell lines (BARHL1, OTX2, NNAT, and degree in the DAOY MB cell line; high MSH1 protein UNC5H; (2) genes that were overexpressed by all the cell levels were observed in four of six human MB tumor lines (TMPO, SET, and SOX4), and (3) genes that specimens. As NUMB is negatively regulated by MSH1 showed variable expression across the cell lines without expression, we studied NUMB expression in the same any distinct pattern (ULIP, ENC1, MSH1 CCND2) samples. NUMB was variably expressed as two isoforms (Figure 2). Interestingly, in the first category, BARHL1, throughout the MB cell lines and tumor specimen. NNAT, and UNC5H, although not expressed in the Interestingly, there was no inverse correlation between tumor cell lines, were expressed in the MB tumor MSH1 and NUMB expression as might be expected in samples suggesting that these genes may be specific to the MB tumor samples.

Figure 2 RT–PCR of genes interest in various human cancer cell lines. Similar to the MB specimens, the expression proﬁle in the array of cell lines could be categorized into three groups: (1) genes that were not primarily expressed by any of the cell lines (BARHL1, OTX2, NNAT, and UNC5H; (2) genes that were overexpressed by all the cell lines (TMPO, SET, and SOX4), and (3) genes that showed variable expression across the cell lines without any distinct pattern (ULIP, ENC1, MSH1 CCND2) b-Actin demonstates equal expression across the panel. hCB, non-neoplastic human cerebellum; HMB, human medulloblastoma case, from which human MB subtractive library was developed, as negative and positive control

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3448

Figure 4 MSH1 expression correlates with increased HES1- luciferase activity. (a) TE671, ONS767, PFSK, UW228, and Figure 3 Western analysis of WNT and Notch pathways in MB UW426 cells were transiently cotransfected with HES1-luciferase specimens and cell lines. The protein expression of the tested genes b-galactosidase reporter constructs. Normalized data for the four is variable across both tumors and cell lines. CCND1 and cell lines show high HES1-luciferase activity in the TE671 cell line CNNTB1 expression levels correlate in the MB cell lines. Notch that expresses high levels of MSH1. Error bars represent standard pathway dysregulation is suggested by the positive correlation deviations. (b) Non-MSH1-expressing cell lines, UW228 and between NUMB and MSH1 expression in the MB tumor UW426, were transfected with empty vector, pEF-MSH1 or specimens. b-Actin levels demonstrates equal protein loading of pcDNA3.1-NICD (Notch1 intracellular domain). The transfection the samples of these cell lines with MSH1 resulted in a 4–6-fold increase in HES1 promoter activity

Hes1 luciferase reporter assays HES1-luciferase activity in TE671 cells was higher by genomic, transcriptomic, and proteomic profiles may 3–-fold compared to the non-MSH1-expressing cell lines give rise to a specific tumor type, it is becoming (Figure 4a). To verify MSH1 regulation of the HES1- increasingly important to map these profiles since each luciferase reporter, we transfected MSH1 into two non- may be associated with responses to different treatment MSH1-expressing MB cell lines, UW228 and UW426. regimens. As such, our study was designed to identify MSH1 expression in these cell lines resulted in a 6–8-fold genes with differential expression between MBs and cells increase in HES1 promoter activity (Figure 4b). Trans- of the non-neoplastic cerebellum. To achieve our goal, fection of these cell lines with the intracellular domain of we used two different models: in the first, we compared NOTCH1, a strong activator of HES1, resulted in a overexpressed genes in a sporadic, classical human MB 40–50-fold induction of HES1 activity as compared to to human non-neoplastic cerebellum; and in the second, the wild-type cell lines. These data suggest that the we took advantage of the HH-driven, murine MB model þ /À Notch pathway may be functional in the UW228 and that arises in Ptch heterozygous knockout mice, and UW426 MB cell lines. compared a MB from these mice to the normal cerebellum of a Ptch þ /À littermate. Interestingly, but perhaps not unexpectedly, we show that the differentially expressed and developmentally regulated genes Discussion from each model, the human sporadic MB and the murine Ptch þ /À MB, are different. Identification of genes that are differentially expressed Several new techniques have recently been developed between cancer cells and nonmalignant cells of the same for analysing altered gene expression. These include phenotype may lead to a greater understanding of which mRNA differential display (Liang and Pardee, 1992), oncogenic processes and signaling pathways are perti- serial analyses of gene expression (SAGE) (Velculescu nent in cellular transformation leading to MB. Human et al., 1997), cDNA microarray (Chee et al., 1996; Chu MB, as with many human cancers, is characterized by et al., 1998), and SSH (Diatchenko et al., 1996). genetic and phenotypic heterogeneity despite a fairly Whereas mRNA differential display is the most uniform appearance by light microscopy. As different complicated and labor-intensive, both SAGE and

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3449 cDNA array are limited for identifying unknown genes We also looked at CCND1 and CCND2 expression by because they a priori require known genetic information. Western blot analysis. Interestingly, while there was SSH is an efficient and versatile PCR-based method of some overlap in expression of CCND1 and CCND2 in identifying rare, tumor-specific transcripts (Ji et al., the MBs tested here, CCND1 was more widely 2002). We previously used SSH to identify differentially expressed in the MB samples than was CCND2. Not expressed genes in an in vivo animal model of cerebral surprisingly, there was good correlation between the ischemia (Yokota et al., 2001). This is the first report of expression of CNNTB1 and CCND1 in the MB cell using SSH on human and murine MB to identify lines, suggesting a potential role of the WNT pathway in differentially expressed genes – many of which have not these cell lines. previously been linked to MB oncogenesis. As with Our isolation of Unc5h4 in murine MB suggests a other PCR-based subtraction methods, some genes were possible role of disrupted apoptosis in this tumor. identified by multiple cDNAs (seen as multiple hits in UNC5H is a membrane netrin-1 receptor crucial for the database search). This could be due to a more axon guidance, neuronal migration, growth cone attrac- abundant transcript resulting in higher cDNA repre- tion, and/or repulsion (Hong et al., 1999). Typically, sentation, or a large transcript leading to multiple UNC5H acts as dependence receptor inducing apoptosis isolated cDNA fragments, or an artefact of PCR when unbound to ligand and may be a tumor suppressor amplification. The number of unknown ESTs in the gene in such tissues as colon, breast, ovary, and uterus. human genome project database has dramatically However, when bound to netrin-1, UNC5H acts as a decreased, but the number in the murine database is potent survival factor (Llambi et al., 2001). UNC5H2 is still quite large and therefore it is not surprising that our highly expressed in cerebellar external granule cells identification rate was higher for human ESTs than along the parallel fibers interacting with Purkinje cell those obtained from the murine library. dendrites (Leonardo et al., 1997). Various UNC5H After differentially expressed MB genes were identi- homologs are known (Leonardo et al., 1997) and human fied, a subset was chosen for further analysis. We chose UNC5H4 (UNC5D) was recently found in the human genes that are developmentally regulated, important for genome database. Another neuronal process-associated cerebellar development or are expressed in the cerebellar protein identified in our human MB library is the Unc- granule cell precursors, the cells that may give rise to 33-like phosphoprotein (ULIP) (Goshima et al., 1995), a MB. Some of these genes direct important cellular developmentally expressed homolog (Byk et al., 1996) of processes such as cell cycle control, apoptosis, and Caenorhabditus elegans Unc-33 (Li et al., 1992), which is phosphorylation, Several are members of well-studied highly expressed between E15 and P15 in rat cerebellum developmental gene families such as the homeobox and (Wang and Strittmatter, 1996). Sry-box gene families, or are components of important An interesting finding in this study was the over- developmental pathways such as the Wnt and Notch expression of Msh1 in the murine MB. MSH1 is an signaling pathways. evolutionarily conserved gene encoding a neural RNA- We identified nine differentially expressed genes from binding protein. In mammals, MSH1 is selectively the murine Ptch þ /À MB and analysed five in more detail. expressed in neural progenitor cells (Okano et al., As the murine cDNA library was generated from a 2002) and is involved in cellular differentiation and ptch þ /À MB in which HH signaling is activated, the tumorigenesis. In stem cells, MSH1 maintains the cells finding of CCND2 upregulation supports the efficacy of in an undifferentiated state through post-transcriptional the technique used. We identified 21 differentially inhibition of NUMB expression (Imai et al., 2001), expressed genes from the classical human MB and which in turn acts as an intracellular Notch antagonist further characterized six genes: ULIP, SOX4, NNAT, and is involved in regulation of neuronal differentiation ENC1, OTX2,andBARHL1. (Wakamatsu et al., 1999). Accordingly, MSH1 is Increased expression of the cell cycle control gene, considered a positive regulator of Notch signaling. CCND2, was detected in our murine Ptch þ /À MB Interestingly, it has been shown that MSH1 is expressed library. Cyclin D, a G1/S cyclin, promotes cells to enter in certain types of brain tumors such as glioblastoma S phase by releasing E2F transcription factors from the and MB, and its expression level correlates with the retinoblastoma protein (Rb) (Weinberg, 1995). CCND2 malignancy and proliferative activity of gliomas (Kane- is transiently expressed in the external granule cells mura et al., 2001). Activation of Notch signaling results during cerebellar ontogenesis (Ross et al., 1996). In mice in tumorigenesis by virtue of inhibition of apoptosis and lacking CCND2, there is a paucity of cerebellar granule maintenance of an immature cell phenotype (Artavanis- cells from a decreased proliferative potential and an Tsakonas et al., 1999). In keeping with MB formation, increased apoptotic rate (Huard et al., 1999). Recently, Notch2 activity has been shown to halt the differentia- CCND2 was shown by microarray analysis to be a tion of cerebellar granule cells and to act as a strong target gene of the zinc-finger transcriptional factor, mitogen (Solecki et al., 2001). As such, high levels of GLI1, and therefore under the regulation of the HH MSH1 expression may lead to increased cellular pathway (Yoon et al., 2002). HH signaling promotes proliferation. Our results indicate a potential dysregula- transcription of CCND2 and cell growth (Duman-Scheel tion of the Notch pathway as MSH1 and NUMB et al., 2002). In our study, the upregulation of Ccnd2 in expression levels were not inversely related in the same the Ptch þ /À MB library, and its high expression in many tumor samples. In support of this statement, the only MB cases support a role for CCND2 in MB oncogenesis. cell line to have high MSH1 expression, TE671, also

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3450 demonstrated high levels of downstream Notch activity desmoplastic MB subtype (Rubin and Rowitch, 2002). as measured by the HES1-luciferase reporter construct. Furthermore, our results suggest the potential dysregu- Furthermore, induced expression of MSH1 in non- lation of the Notch pathway in MB. The identification MSH1-expressing cell lines leads to a 6–8-fold induction of additional MB-associated markers may not only of HES1 activity. This induction suggests that the improve MB tumor grading but may also predict for intracellular Notch signaling pathway is intact in these patients who will respond best to a given therapy. In MB cell lines. future studies, we will examine the panel of genes The SOX (Sry-box) genes are developmentally regu- identified here and their pattern of expression against lated and homologous to the HMG-box region of the large numbers of patient samples linked to clinical testis-determining gene, SRY (Laudet et al., 1993). outcome and survival data. SOX4 is expressed in mouse T- and pre-B-lymphocytic cell lines along with LEF-1, and controls lymphocyte differentiation (van de Wetering et al., 1993). In the Materials and methods developing cerebellum, SOX4 is expressed in the external granule cells (Cheung et al., 2000). Interest- Cell lines and culture conditions ingly, Sox4 expression was repressed in stable Gli1- et al MB cell lines TE671, ONS76, UW228, UW426, PFSK, transfected rat kidney epithelial cells (Yoon ., 2002). DAOY, and D283 were gifts of Dan Fults, the University of A downstream effector of Wnt signaling along with Utah, Salt Lake city, and cultured in Dulbecco’s modified LEF-1, SOX4 may play a role in tumor cell differentia- Eagle’s medium (DMEM) supplemented with 10% (vol/vol) tion. SOX4 is known to be overexpressed in MB, fetal bovine serum (FBS, Wisent, St Bruno, Quebec, Canada), especially the classic variant and has been suggested as a 200 mML-glutamine, 5 U of penicillin/ml, and 5 mg of favorable prognostic factor (Lee et al., 2002). streptomycin sulfate/ml. D341 cells were grown in DMEM Other identified genes from the MB libraries are supplemented with 20% FBS. The other cell lines used in this important in cerebellar development. The homeobox study were astrocytoma cell lines (SF188, U87, SF539, U251, proteins (BARH1, OTX2) and NNAT exhibit neural- SF126, U343, and U373), and other human cancer cell lines specific expression throughout early development. (Jurkat, T-cell lymphoma; Ramos, Burkitt’s lymphoma; HeLa, squamous cell carcinoma; SA052, osteosarcoma; U205, BARHL1 Mammalian is expressed in the rhombence- osteosarcoma; COLO320DM, adenocarcinoma; NCI-173, phalon, including the cerebellar external granule cells small-cell carcinoma; and NCI-H596, adeno-squamous carci- and spinal dorsal sensory interneurons, and required for noma). These cell lines have all been well characterized and the generation of specific neuronal subtypes (Bulfone were purchased from American Type Culture Collection et al., 2000). BARHL1 is a candidate gene for Joubert (ATCC, Manassas, VA, USA). syndrome, a rare developmental disorder characterized by aplasia/hypoplasia of the cerebellar vermis, possibly Origin of MB cDNA libraries playing an important role in cerebellar ontogenesis. Mouse BARHL1 activates the proneural gene neuro- A human MB subtractive cDNA library was developed from a genin2 (ngn2) and inhibits MASH1 (Saito et al., 1998). tumor specimen of a 5-year-old female subject diagnosed with MB at The Hospital for Sick Children, Toronto. Histopatho- Human BARHL1 has been mapped to chromosome logical examination showed a classical MB containing small 9q34 (Bulfone et al., 2000), adjacent to PTCH (9q22.3), round tumor cells with abundant chromatin and scanty, SET (9q34.11), and NOTCH1 (9q34.3) – an important poorly defined cytoplasm. Control mRNA was derived from locus known to be deleted in MBs (Bayani et al., 2000). non-neoplastic human cerebellum and purchased from Clon- Another identified homeobox protein, mammalian tech (Pale Alto, CA, USA). Otx2, is expressed in the telencephalon, diencephalon, A murine MB subtractive cDNA library was derived from and mesencephalon of developing mouse brain embryos. specimens of Ptch þ /À heterozygous mice. Control mRNA for (Boncinelli et al., 1993). In the cerebellum, Otx2 is this subtraction was generated from normal cerebellum of þ /À expressed in the entire cerebellar granule cell lineage unaffected littermates. Ptch heterozygous mice were a gift (Frantz et al., 1994). Recently, OTX2 and ZIC1 were from CC Hui, The Hospital for Sick Children. Mice genotyping was performed by PCR on DNA isolated from identified as differentially expressed genes in MB by À/À tail specimens as described previously (Goodrich et al., 1997). SAGE (Michiels et al., 1999). Otx2 mice are Permission to use the above biological materials was granted embryonic lethal showing defective development of the by the Research Ethics Board and the Committee of the rostral neuroectoderm resulting in a headless phenotype Animal Care Facility of The Hospital for Sick Children. (Simeone et al., 1992). In summary, several different patterns of gene PCR-based suppression subtractive hybridization expression were found in our survey of MB tumor samples. The heterogeneity of MB tumors was revealed Total RNA was prepared with TRIZOL (Invitrogen Canada by testing with multiple markers important for cerebel- Inc., Burlington, Canada). mRNA was then isolated with lar development that were derived from our SSH screen. Oligotex (QIAGEN Inc., Mississauga, Canada). Double- stranded cDNA (ds cDNA) was synthesized by the Gubler– This suggests that MB is not a uniform tumor, but may Hoffman method using a PCR-Select cDNA Subtraction Kit be comprised of several possible molecular entities as (Clontech, Palo Alto, CA, USA). For the murine Ptch þ /À illustrated by the patterns of gene expression found in library, the amount of mRNA retrieved was limited such that the MB samples and cell lines examined here. Several cDNA amplification was performed using a SMART PCR reports implicate the WNT and HH pathways in the cDNA Synthesis Kit (Clontech, Palo Alto, CA, USA).

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3451 Subtractive cDNA libraries for the human and murine QIAEX II (QIAGEN Inc., Mississauga, Canada), and materials were established by SSH as previously described sequenced by the dideoxy chain termination method using a (Diatchenko et al., 1996; Yokota et al., 2001). Briefly, ds DNA autosequencer, 373A/GeneScanTM (PE Biosystem, cDNA was digested with the blunt-ended restriction enzyme Foster City, CA, USA). DNA sequences were analysed by RsaI. Tester ds cDNA was synthesized from the MB tumor the DNASIS program (Hitachi, Japan), and a homology tissue, and driver ds cDNA from normal cerebellum. Follow- search was performed against the sequences in the database ing heat denaturation (1.5 min at 981C), the digested tester ds using the NCBI BLAST program (http://www.ncbi.nih. cDNA was ligated to either adapter 1: (50-CTAATACGACT- gov:80/BLAST/). CACACTATAGGGCTCGAGCGGCCGCCGGGCAGGTA- CCTGCCCGG-30) or adapter 2R: (5-CTAATACGACTCA- RT–PCR in MB specimens and cell lines CACTATAGGGCAGCGTGGTCGCGGCCGAGGTACC- TCGGCCG-30), in separate reactions. They were then mixed RT–PCR analyses were performed for developmentally with an excess of driver cDNA and hybridized (50 mM regulated genes from the SSH experiments in 25 cases of HEPES, pH 8.3, 0.5 M NaCl, 0.2 mM EDTA) for 10 h at sporadic MB and 19 human cancer cell lines (4 MB cell lines 681C. After the first hybridization, these two samples were (Uw228, DAOY, PFSK, Med344); seven glioma cell lines mixed, and the heat-denatured driver cDNA was added to the (SF188, U87, SF539, U251, SF126, U343, U373); and other same hybridization buffer. The hybridization was allowed to human cancer cell lines (Jurkat, Ramos, HeLa, SA052, U205, proceed for an additional 10 h at 681C. The final hybridization COLO320DM, NCI-173, and NCI-H596)). mixture was diluted in dilution buffer (20 mM HEPES, pH 8.3, mRNA was prepared as described above. A 5 mg measure of 50 mM NaCl, 0.2 mM EDTA), heated at 721C for 7 min and the total RNA prepared from normal cerebellum, MBs, and stored at À201C. PCR amplification was performed using the human cancer cell lines were digested with DNaseI, and used Advantage cDNA PCR Core Kit (Clontech, Palo Alto, CA, as templates for cDNA synthesis with the oligo (dT) USA). The subtracted cDNA was diluted and the first PCR (Sambrook et al., 1989; Liang and Pardee, 1992; Velculescu was performed using the PCR primers P1 (50- CTAATAC- et al., 1997; Yokota et al., 2001; Ji et al., 2002) primers using GACTCACACTATAGGGC-30) using the following protocol: the Omniscript cDNA synthesis Kit (Qiagen Inc., Mississauga, 751C for 7 min, followed by 30 cycles at 941C for 30 s, 681C for Canada). The reaction mixture was incubated at 371C for 2 h, 30 s, and 721C for 150 s, and final extension at 681C for 7 min. and then terminated by heating at 931C for 5 min. The The amplified product was diluted and secondary PCR was resulting cDNA mixture was diluted in 100 ml of 1/10 TE (pH performed under the same conditions for 10 cycles using the 8.0) before PCR amplification. PCR amplification was nested primers PN1 (50-TCGAGCGGCCGCCGGGCAGGT- performed by adding a 2-ml aliquot of each cDNA sample to 30) and PN2R (50-AGCGTGGTCGCGGCCGAGG-T-30). 25 ml of the reaction mix from the Advantage cDNA PCR Products from the secondary PCR were inserted into pBlue- Core Kit (Clontech, Palo Alto, CA, USA). The gene-specific script II (Stratagene, La Jolla, CA, USA) at the EcoRV site primers used in this study were designed from the known and transformed into competent cells, Escherichia coli DH5a. sequences of the isolated genes shown in Table 1. The (Invitrogen, Burlington, Canada). following amplification protocol was used: denaturation at 951C for 30 s, annealing at each temperature (shown in Table 1) for 30 s, and extension at 721C for 1 min. After PCR, 8 mlof Isolation of upregulated cDNA fragments and their identification each of the PCR products were electrophoretically separated Randomly selected bacterial colonies were selected and on 2% agarose gel and stained with ethidium bromide. amplified under the same conditions for 20 cycles with the Densitometric analyses were performed including standardiza- nested primers 1 and 2R described above. The PCR products tion from the control (b-actin) band as previously described were blotted onto GeneScreen Plus (MEN Life Science (Yokota et al., 2001). The rate of amplification with the primer Products Inc., Boston, MA, USA), denatured twice in 0.5 N sets used here was exponential for each cycle determined, and NaOH on Whatman 3 MM filter paper for 2 min, then the pattern of amplification remained constant through neutralized by dipping in Tris-HCl (pH 7.5) for 2 min and different runs. finally were fixed to the filter by ultraviolet light irradiation. The forward- and reverse-subtracted secondary PCR products Western blot analysis of MB specimens and cell lines were purified using QIAquick Spin (QIAGEN Inc., Mississau- ga, Canada) and labeled with 32P ATP (Amersham Pharmacia MB tumour specimens and cell lines were lysed in PLC lysis Biotech, Piscataway, NJ, USA) using a Random Primed DNA buffer (50 mM HEPES (pH 7.5), 150 mM NaCl, 10% (vol/vol) Labeling Kit (Roche, Laval, Quebec, Canada). The filters were glycerol, 1% (vol/vol) Triton X-100, 1.5 mM MgCl2,1mM air-dried and prehybridized in ExpressHyb Hybridization EDTA, 10 mM NaPPI, 100 mM NaF) containing complete Solution (Amersham Pharmacia Biotech, Piscataway, NJ, protease inhibitors (Roche, Laval, Quebec). Lysates were USA) with herring sperm DNA, oligonucleotides of nested cleared by centrifugation at 14 000 r.p.m. at 41C for 10 min. A primers, and their complementary sequences for 1 h. Hybridi- measure of 40 mgof total protein lysates were resolved on 10% zation was allowed to proceed at 681C overnight after the SDS–PAGE gels and electrophoretically transferred to poly- addition of 106 c.p.m./ml of 32P-labeled cDNA probes prepared vinylidene fluoride (PVDF) membranes (Millipore, Bedford, from either the MB, or the non-neoplastic cerebellum. The MA, USA). Membranes were blocked with 5% (wt/vol) skim filters were washed twice with 2 Â SSC and 0.1 Â SSC with milk powder in 1 Â Tris-buffered saline with 0.05% Tween 20 0.5% SDS at 651C and exposed to an imaging plate and for 30 min prior to the antibody addition. Blocked membranes analysed by STORM gel and blot imaging system (Amersham were incubated for 1 h at room temperature with rat anti- Bioscience, Sunnyvale CA, USA). human MSH-1 antibody (a kind gift from Professor H Okano, The clones that were positive for the subtracted probe and Dokkyo University, Tochigi, Japan) at a dilution of 1 : 1000, negative for the reverse-subtracted one were selected and their rat anti-human cyclin D2 antibody (Santa Cruz Biotechnol- plasmids were prepared once again. The method for plasmid ogy, Santa Cruz, CA, USA) at a dilution of 1 : 1000, anti- DNA preparation has been described previously (Sambrook human actin antibody (Sigma, The Woodlands, TX, USA) at a et al., 1989). The desired cDNAs were then purified using dilution of 1 : 1000, mouse anti-human cyclin D1 (Santa Cruz

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3452 Biotechnology, Santa Cruz, CA, USA) at a dilution of 1 : 1000, McGlade, 2003). Each transfection was performed in duplicate rabbit anti-numb (UBI, Waltham, MA, USA), and rabbit anti- and each experiment was conducted four times. b-catenin (Santa Cruz Biotechnology, Santa Cruz, CA, USA) UW228 and UW426 cells were seeded on six-well plates and at a dilution of 1 : 1000. Membranes were incubated with transiently transfected with 1 mg of HES1-luciferase reporter horseradish peroxidase (HRP)-conjugated secondary antibody construct, 1 mgofb-galactosidase reporter construct, and (BioRad, Hercules, CA, USA) at room temperature for 45 min either 1 mg of empty vector, 1 mg pEF-MSH1 expression at a dilution of 1 : 5000. Bound antibody was detected using construct, or pcDNA3.1-NICD (NOTCH1 intracellular do- enhanced chemiluminescence reagent (ECL, PerkinElmer, main) using FUGENE 6 at a 3 : 1 ratio. Luciferase and b- Boston, MA, USA). galactosidase assays were performed 24 h post-transfection. Each transfection was carried out in triplicate and each Hes1 luciferase reporter assays experiment was conducted three times. A MSH1-expressing cell line (TE671) and cell lines that do not express MSH1 (ONS76, PFSK, UW228, UW426) were Acknowledgements assayed for Notch signaling activity. TE671, ONS76, PFSK, This study was supported in part by grants from the UW228, and UW426 cells were seeded on six-well plates and National Cancer Institute of Canada, the Pediatric Brain transiently transfected with 1 mgofHES1-luciferase reporter Tumor Foundation of the United States, the Ontario Cancer construct (a gift from Jane McGlade, Toronto, Ontario, Research Network and the Neurosurgical Research and Canada) using Fugene 6 transfection reagent (Roche, Laval, Educational Foundation. Dr Rutka is a Scientist of the Quebec, Canada). b-Galactosidase reporter construct (1 mg) Canadian Institutes of Health Research. Dr Yokota was was included in each transfection to control for efﬁciency. partially supported by a grant from the Japanese Ministry Luciferase and b-galactosidase assays were performed 24 h of Education, Culture, Sports, Science and Technology post-transfection as previously described (McGill and (No 13671430).

References

Artavanis-Tsakonas S, Rand MD and Lake RJ. (1999). Ji W, Wright MB, Cai L, Flament A and Lindpaintner K. Science, 284, 770–776. (2002). BMC Genom., 3, 12. Bayani J, Zielenska M, Marrano P, Kwan Ng Y, Taylor MD, Kanemura Y, Mori K, Sakakibara S, Fujikawa H, Hayashi H, Jay V, Rutka JT and Squire JA. (2000). J. Neurosurg., 93, Nakano A, Matsumoto T, Tamura K, Imai T, Ohnishi T, 437–448. Fushiki S, Nakamura Y, Yamasaki M, Okano H and Arita Boncinelli E, Gulisano M and Broccoli V. (1993). N. (2001). Differentiation, 68, 141–152. J. Neurobiol., 24, 1356–1366. Kozmik Z, Sure U, Ruedi D, Busslinger M and Aguzzi A. Buhren J, Christoph AH, Buslei R, Albrecht S, Wiestler OD (1995). Proc. Natl. Acad. Sci. USA, 92, 5709–5713. and Pietsch T. (2000). J. Neuropathol. Exp. Neurol., 59, Laudet V, Stehelin D and Clevers H. (1993). Nucleic Acids 229–240. Res., 21, 2493–2501. Bulfone A, Menguzzato E, Broccoli V, Marchitiello A, Lee CJ, Appleby VJ, Orme AT, Chan WI and Scotting PJ. Gattuso C, Mariani M, Consalez GG, Martinez S, Ballabio (2002). J. Neurooncol., 57, 201–214. A and Banﬁ S. (2000). Hum. Mol. Genet., 9, 1443–1452. Leonardo ED, Hinck L, Masu M, Keino-Masu K, Byk T, Dobransky T, Cifuentes-Diaz C and Sobel A. (1996). Ackerman SL and Tessier-Lavigne M. (1997). Nature, 386, J. Neurosci., 16, 688–701. 833–838. Chee M, Yang R, Hubbell E, Berno A, Huang XC, Stern D, Li W, Herman RK and Shaw JE. (1992). Genetics, 132, Winkler J, Lockhart DJ, Morris MS and Fodor SP. (1996). 675–689. Science, 274, 610–614. Liang P and Pardee AB. (1992). Science, 257, 967–971. Cheung M, Abu-Elmagd M, Clevers H and Scotting PJ. Llambi F, Causeret F, Bloch-Gallego E and Mehlen P. (2001). (2000). Brain Res. Mol. Brain Res., 79, 180–191. EMBO J., 20, 2715–2722. Chu S, DeRisi J, Eisen M, Mulholland J, Botstein D, Brown McGill MA and McGlade CJ. (2003). J. Biol. Chem., 278, PO and Herskowitz I. (1998). Science, 282, 699–705. 23196–23203. Diatchenko L, Lau YF, Campbell AP, Chenchik A, Moqadam Michiels EM, Oussoren E, Van Groenigen M, Pauws E, F, Huang B, Lukyanov S, Lukyanov K, Gurskaya N, Bossuyt PM, Voute PA and Baas F. (1999). Physiol. Genom., Sverdlov ED and Siebert PD. (1996). Proc. Natl. Acad. Sci., 1, 83–91. USA, 93, 6025–6030. Okano H, Imai T and Okabe M. (2002). J. Cell Sci., 115, Duman-Scheel M, Weng L, Xin S and Du W. (2002). Nature, 1355–1359. 417, 299–304. Pomeroy SL, Tamayo P, Gaasenbeek M, Sturla LM, Angelo Frantz GD, Weimann JM, Levin ME and McConnell SK. M, McLaughlin ME, Kim JY, Goumnerova LC, Black PM, (1994). J. Neurosci., 14, 5725–5740. Lau C, Allen JC, Zagzag D, Olson JM, Curran T, Wetmore Goodrich LV, Milenkovic L, Higgins KM and Scott MP. C, Biegel JA, Poggio T, Mukherjee S, Rifkin R, Califano A, (1997). Science, 277, 1109–1113. Stolovitzky G, Louis DN, Mesirov JP, Lander ES and Goshima Y, Nakamura F, Strittmatter P and Strittmatter SM. Golub TR. (2002). Nature, 415, 436–442. (1995). Nature, 376, 509–514. Ross ME, Carter ML and Lee JH. (1996). J. Neurosci., 16, Hong K, Hinck L, Nishiyama M, Poo MM, Tessier-Lavigne 210–219. M and Stein E. (1999). Cell, 97, 927–941. Rostomily RC, Bermingham-McDonogh O, Berger MS, Huard JM, Forster CC, Carter ML, Sicinski P and Ross ME. Tapscott SJ, Reh TA and Olson JM. (1997). Cancer Res., (1999). Development, 126, 1927–1935. 57, 3526–3531. Imai T, Tokunaga A, Yoshida T, Hashimoto M, Mikoshiba Rubin JB and Rowitch DH. (2002). Cancer Cell, 2, 7–8. K, Weinmaster G, Nakafuku M and Okano H. (2001). Mol. Saito T, Sawamoto K, Okano H, Anderson DJ and Mikoshiba Cell Biol., 21, 3888–3900. K. (1998). Dev. Biol., 199, 216–225.

Oncogene Developmentally regulated genes in medulloblastoma N Yokota et al 3453 Sambrook J, Fritsch E and Maniatis T. (1989). Molecular Wang LH and Strittmatter SM. (1996). J. Neurosci., 16, Cloning, 2nd edn. Cold Springs Harbor Laboratory Press: 6197–6207. Cold Springs Harbor. Wechsler-Reya R and Scott MP. (2001). Annu. Rev. Neurosci., Simeone A, Acampora D, Gulisano M, Stornaiuolo A and 24, 385–428. Boncinelli E. (1992). Nature, 358, 687–690. Weinberg RA. (1995). Cell, 81, 323–330. Solecki DJ, Liu XL, Tomoda T, Fang Y and Hatten ME. Wetmore C, Eberhart DE and Curran T. (2001). Cancer Res., (2001). Neuron, 31, 557–568. 61, 513–516. Taylor MD, Mainprize TG and Rutka JT. (2000). Neurosur- Yokota N, Aruga J, Takai S, Yamada K, Hamazaki M, Iwase gery, 47, 888–901. T, Sugimura H and Mikoshiba K. (1996). Cancer Res., 56, van de Wetering M, Oosterwegel M, van Norren K and 377–383. Clevers H. (1993). EMBO J., 12, 3847–3854. Yokota N, Nishizawa S, Ohta S, Date H, Sugimura H, Velculescu VE, Zhang L, Zhou W, Vogelstein J, Basrai MA, Namba H and Maekawa M. (2002). Int. J. Cancer, 101, Bassett Jr DE, Hieter P, Vogelstein B and Kinzler KW. 198–201. (1997). Cell, 88, 243–251. Yokota N, Uchijima M, Nishizawa S, Namba H and Koide Y. Vincent S, Turque N, Plaza S, Dhellemmes P, Hladky JP, (2001). Stroke, 32, 168–174. Assaker R, Ruchoux MM and Saule S. (1996). Int. J. Yoon JW, Kita Y, Frank DJ, Majewski RR, Cancer, 8, 901. Konicek BA, Nobrega MA, Jacob H, Walterhouse D Wakamatsu Y, Maynard TM, Jones SU and Weston JA. and Iannaccone P. (2002). J. Biol. Chem., 277, (1999). Neuron, 23, 71–81. 5548–5555.

Oncogene