Oncogene (2010) 29, 3017–3024 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 $32.00 www.nature.com/onc SHORT COMMUNICATION The SFRP family of WNT inhibitors function as novel tumor suppressor epigenetically silenced in medulloblastoma

PN Kongkham1,2, PA Northcott1,3, SE Croul4, CA Smith1,2, MD Taylor1,3 and JT Rutka1,2

1Division of Neurosurgery, Arthur and Sonia Labatt Brain Tumor Research Centre, Toronto, Ontario, Canada; 2Program in Cell Biology, Hospital for Sick Children, University of Toronto, Toronto, Ontario, Canada; 3Program in Developmental and Stem Cell Biology, Hospital for Sick Children, University of Toronto, Toronto, Ontario, Canada and 4Department of Pathology, University Health Network, University of Toronto, Toronto, Ontario, Canada

Medulloblastoma (MB) is the most common malignant signaling is crucial for normal cerebellar development, pediatric brain tumor. Dysregulation of WNT signaling as transgenic mice with reduced WNT signaling exhibit occurs in up to 20% of cases. Using a genome-wide cerebellar aplasia (McMahon and Bradley, 1990). approach, we identified the secreted frizzled-related Turcot syndrome patients with germline adenomatous 1, 2 and 3 (SFRP1, SFRP2 and SFRP3) family polyposis coli (APC) and resultant of WNT inhibitors as putative tumor suppressor genes increased WNT signaling have an increased risk of silenced by promoter region methylation in MB. SFRP1, MB (Hamilton et al., 1995). Furthermore, SFRP2 and SFRP3 expression increased after 5-aza-20- affecting WNT signaling members (APC, CTNNB1, deoxycytidine treatment. SFRP1, SFRP2 and SFRP3 AXIN1 and AXIN2) are seen in 15–20% of sporadic MB methylation was identified in 23.5, 3.9 and 15.7% of cases (Eberhart et al., 2000; Baeza et al., 2003; primary MB specimens, respectively, by methylation- Thompson et al., 2006; Koch et al., 2007). specific PCR. Stable SFRP1, SFRP2 and SFRP3 Recent gene expression profiling studies show that expression reduced phospho-DVL2 levels and hindered human MBs may be divided into distinct molecular MB cell proliferation and colony formation in soft agar subgroups including one characterized by expression of in vitro. In 60% of primary tumors, SFRP1 was expressed markers of canonical WNT signaling such as WIF1, at levels twofold lower than that in normal cerebellum. DKK2, LEF1, KREMEN1 and AXIN2 (Thompson SFRP1 expression impaired tumor formation in vivo in et al., 2006; Kool et al., 2008; Northcott et al., 2009a). flank and orthotopic intracerebellar xenograft models and Most tumors in the WNT subgroup have monosomy 6, conferred a significant survival advantage (Po0.0001). activating CTNNB1 mutations, and carry a more We identify for the first time favorable prognosis compared with other MB sub- function of SFRP genes in MB, and suggest that loss of groups (Eberhart et al., 2000; Ellison et al., 2005; Kool WNT pathway inhibition due to SFRP gene silencing is an et al., 2008; Northcott et al., 2009a). additional mechanism that may contribute to excessive Recently, we used a microarray-based genome-wide WNT signaling in this disease. epigenetic approach to identify novel tumor suppressor Oncogene (2010) 29, 3017–3024; doi:10.1038/onc.2010.32; genes silenced by promoter methylation in MB (Kongk- published online 8 March 2010 ham et al., 2008). Among candidates identified were members of the secreted frizzled-related protein (SFRP) Keywords: medulloblastoma; WNT signaling; epige- family of WNT inhibitors, including SFRP1, SFRP2 netics; methylation; tumor suppressor gene and SFRP3. Functionally, SFRP genes can limit canonical and noncanonical WNT signaling, binding WNT ligand and sequestering it away from Fz receptors (Bovolenta et al., 2008). SFRPs have been identified as tumor suppressor genes whose expression is reduced in a Introduction variety of malignancies (Dahl et al., 2007). Only SFRP1 has been examined in the context of MB previously. Medulloblastoma (MB) is the most common malignant SFRP1 expression has been shown to increase after posterior fossa tumor. Dysregulation of developmental treatment of the D283 cell line with the histone signaling pathways, including WNT signaling, is known deacetylase inhibitor, trichostatin, but the functional to have a major role in MB pathogenesis. WNT significance of this was not examined (Vibhakar et al., 2007). A second study looking at SFRP1 in MB failed to Correspondence: Dr JT Rutka, The Division of Neurosurgery, Arthur identify promoter methylation, which may have been in and Sonia Labatt Brain Tumor Research Centre, Suite 1503, The part due to the limited number of tumors examined Hospital for Sick Children, 555 University Avenue, Toronto, Ontario, (Chang et al., 2005). Canada M5G 1X8. E-mail: [email protected] Here, we show for the first time that SFRP1, SFRP2 Received 1 September 2009; revised 15 November 2009; accepted 3 and SFRP3 promoter region methylation is associated January 2010; published online 8 March 2010 with gene silencing in MB cell lines and primary tumors. SFRP gene silencing in medulloblastoma PN Kongkham et al 3018 Expression of SFRP1, SFRP2 or SFRP3 in MB cells reduced phospho-DVL2 in both the D283 and ONS76 reduces their proliferative capacity and anchorage- cell lines compared with controls (Figure 2a and Supple- independent growth in vitro. SFRP1, SFRP2 or SFRP3 mentary Figure 5). Stable re-expression of SFRP1-3 in expression limits WNT signaling, as evidenced by D283 cells did not alter the levels of activated b-catenin reduced levels of phospho-DVL2. SFRP1 expression (Supplementary Figure 8). Expression of SFRP1, SFRP2 also limits tumor burden and prolongs overall survival and SFRP3 also decreased D283 and ONS76 cell proli- in vivo in an intracranial xenograft model of MB. feration, as measured by MTS (3-(4,5-dimethylthiazol-2-yl)- 2,5-diphenyltetrazolium bromide) assay (Figure 2b and Supplementary Figure 6). Finally, SFRP1, SFRP2 and SFRP3 expression significantly impaired D283 capacity for Results and discussion anchorage-independent growth (Figure 2c and Supplemen- tary Figure 7). SFRP1, SFRP2 and SFRP3 silencing is associated with promoter region methylation Increased expression of SFRP1, SFRP2 and SFRP3 0 MB subgroup-specific expression of SFRP1 after 5-aza-2 -deoxycytidine (5-aza-dC) treatment in MB Using Affymetrix Genechip Human Exon Arrays cell lines was validated using quantitative real-time (Affymetrix, Santa Clara, CA, USA) to assess gene PCR. SFRP1 expression increased in the ONS76, expression in a large cohort of 103 primary MBs, 7 MB UW228, UW426, D425, D458 and MED8A cell lines cell lines, 9 fetal and 5 adult cerebellar samples, we after treatment (Figure 1a). SFRP2 expression increased observed SFRP1 expression levels to be more than in the UW228 cell line, and SFRP3 in the DAOY, D283 twofold lower in 60% (62/103) of primary MBs and MED8A cell lines (Supplementary Figure 1). compared with normal cerebellum. Unsupervised hier- Bisulfite sequencing confirmed promoter region methyl- archical clustering based on 1450 differentially expressed ation for SFRP1, SFRP2 and SFRP3 in MB cell lines genes divided primary tumors into four molecular compared with normal fetal and adult cerebellum subgroups (WNT, Sonic hedgehog (SHH) homolog, (Figure 1b and Supplementary Figures 2 and 3). Methyla- Group C and Group D) as described by our group and tion-specific PCR identified methylation of the SFRP1, others (Thompson et al., 2006; Kool et al., 2008; SFRP2 and SFRP3 promoter regions in 12/51 (23.5%), 2/ Northcott et al., 2009a). Interestingly, SFRP1 expres- 51 (3.9%) and 8/51 (15.7%) primary MB specimens, sion was reduced in WNT, Group C and Group D MB respectively (Figure 1c and Supplementary Figure 4). subgroups in comparison with the SHH group of Where sufficient mRNA was available, SFRP1 expression tumors (Figure 3a). was correlated with SFRP1 promoter methylation status A common feature of WNT subgroup tumors is the by quantitative real-time PCR. Samples identified as having presence of downstream activating CTNNB1 mutations. aberrant promoter region methylation by methylation- One might question the functional significance of specific PCR showed low levels of SFRP1 gene expression reduced SFRP1 expression in the context of downstream compared with normal cerebellum (Figure 1d). activating mutations. However, it has been shown in the context of that loss of SFRP gene SFRP1, SFRP2 and SFRP3 re-expression reduces WNT expression in the presence of downstream mutations of signaling, cell proliferation and anchorage-independent APC or activating CTNNB1 mutations function in a growth synergistic fashion, and that expression of SFRPs can To determine the functional significance of SFRP1, reduce WNT signaling even in the presence of down- SFRP2 and SFRP3 silencing in MB, we stably expressed stream activating mutations (Suzuki et al., 2004). each gene individually in the D283 and ONS76 MB cell Interestingly, SFRP1 expression was also reduced in lines and assessed the effect on WNT signaling, cell Group C and Group D tumors that do not possess proliferation and colony formation in soft agar com- canonical WNT gene expression signatures (Figure 3a). pared with both empty vector and an EGFP-expressing This suggests the possibility that lack of SFRP1 control stable cell lines (Figure 2 and Supplementary expression in these tumors may influence MB pathogen- Figures 5–8). esis through dysregulation of noncanonical WNT SFRPs exert their effect upstream at the level of the signaling pathways. As mentioned, MB cell lines, WNT ligand–Fz receptor interaction, and may therefore including D283 and ONS76, cluster separately from influence both canonical and/or noncanonical WNT the canonical WNT MB subgroup based on Exon Array signaling downstream. In addition, MB cell lines gene expression data (Northcott et al., 2009a). In (including D283 and ONS76) do not possess a canonical addition, they lack the cytogenetic marker of monosomy WNT signaling expression profile or cytogenetic abbera- 6 associated with canonical WNT tumors, but rather tions such as monosomy 6, characteristics typical of possess an isochromosome 17q and either gain or WNT subgroup MBs (Northcott et al., 2009a, b). amplification of MYC—markers associated with Group Therefore, to determine the effect of SFRP expression C and Group D tumors (Northcott et al., 2009b). on WNT signaling, we examined levels of phospho- In contrast, the SHH MB subgroup continues to DVL2, upstream of the point of divergence into the express SFRP1 (Figure 3a). Maintained SFRP1 expres- canonical and noncanonical pathways (Habas and sion in the SHH subgroup may be owing to the fact that Dawid, 2005). SFRP1, SFRP2 and SFRP3 expression SFRP1 is a known SHH signaling target (Katoh and

Oncogene SFRP gene silencing in medulloblastoma PN Kongkham et al 3019

Figure 1 (a) qRT–PCR for SFRP1 expression in MB cell lines before (À)andafter(þ )72htreatmentwith5mM 5-aza-20-deoxycytidine (5-aza-dC). Treatment leads to increased SFRP1 expression in the ONS76, UW228, UW426, D425, D458 and MED8A cell lines. (b) Bisulfite sequencing for SFRP1 showing methylation (filled circles) in ONS76 compared with the unmethylated state (open circles) in both normal fetal and adult cerebellum (Cb). Each row represents experimental replicates. (c) Representative methylation-specific PCR (MSP) data for SFRP1 in the ONS76 cell line, normal cerebellum and primary MB samples (U, unmethylated-specific primers, M, methylated-specific primers). (d) qRT–PCR correlation of SFRP1 expression with methylation status in select MB specimens. Samples known to be methylated, as identified by MSP, show low-level SFRP1 expression compared with normal fetal and adult cerebellum. Representative SFRP1 expression by an unmethylated MB specimen is included for comparison.

Oncogene SFRP gene silencing in medulloblastoma PN Kongkham et al 3020

Figure 2 (a) Western blot for DVL2 in D283 and ONS76 stable transfectants. SFRP1 expression reduced phospho-DVL2 compared with empty vector and EGFP-expressing controls. Anti-DVL2 antibody from Cell Signaling Technologies (Boston, MA, USA). Anti- FLAG antibody from Sigma-Aldrich Inc (Saint Louis, MO, USA). A blot using anti-human transferrin receptor antibody from Invitrogen/Zymed (Burlington, ON, Canada) was used as a loading control. (b) MTS proliferation assays (Promega, Madison, WI, USA) for stably transfected D283 and ONS76 MB cells showing reduced proliferation of two stable D283 clones and one stable ONS76 clone expressing SFRP1, in comparison with empty vector and EGFP-expressing controls. (c) SFRP1 expression in two stable D283 clones significantly reduced their ability to form colonies in soft agar compared with either empty vector-transfected or EGFP- expressing controls (Po0.005 compared with EGFP-expressing controls).

Katoh, 2006). SFRP1 expression is seen in both human population of precursor cells (Gilbertson and Ellison, MB and murine models of MB in which SHH signaling is 2008). In contrast, tumors of the WNT, Group C and upregulated (Lee et al., 2003; Thompson et al., 2006; Kool Group D subtypes may arise from distinct precursor et al., 2008). An important question is whether or not cells residing in the cerebellar subventricular zone or SFRP1 exerts a tumor suppressor gene effect on SHH- white matter, and thus may represent cell populations driven tumors. It is possible that SHH-driven tumors susceptible to alterations in WNT signaling activity (Fan tolerate high levels of SFRP1 expression without suffering and Eberhart, 2008; Gilbertson and Ellison, 2008). deleterious effects from reduced WNT signaling activity. Another possibility is that SHH subgroup tumors The SHH MB subgroup is believed to originate remain susceptible to the tumor suppressor effects of from cerebellar granule neuron precursor cells (Fan SFRP1, and are kept in check by its expression. As and Eberhart, 2008; Gilbertson and Ellison, 2008). existing MB cell lines do not faithfully represent this MB WNT signaling has not been shown to influence the subgroup, determining the functional significance of proliferative capacity or hedgehog responsiveness of this SFRP1 expression in SHH-driven tumors may require

Oncogene SFRP gene silencing in medulloblastoma PN Kongkham et al 3021

Figure 3 (a) Boxplot of Exon Array data for SFRP1 expression in normal cerebellum, primary MB (divided into molecular subgroups) and MB cell lines. SFRP1 is expressed at lower levels in MB cell lines and the WNT, Group C and Group D MBs in comparison with SHH subgroup tumors. (b) Representation of losses encompassing the SFRP1 locus on 8p, as determined by SNP array analysis. (c) Boxplot showing a trend toward reduced SFRP1 expression by exon array in tumors with 8p loss. (d) Boxplot showing the association between high MYC expression and reduced SFRP1 expression by Exon Array in primary MB samples. 500 K Affymetrix SNP GeneChip Mapping Arrays and Affymetrix Genechip Human Exon 1.0 ST Arrays were performed as previously described (Northcott et al., 2009a, b). further study using hedgehog-driven murine models of was identified in only 23.5%, we sought to determine MB or granule neuron precursor cells derived from such whether additional mechanisms may target SFRP1 models. for silencing. The SFRP1 gene locus on chromosome 8p is a frequent target for loss of heterozygosity in a variety of malignancies including MB (Northcott Mechanisms contributing to low-level SFRP1 expression et al., 2009b). We previously reported results from a As SFRP1 expression was twofold lower in 60% high-resolution single-nucleotide polymorphism array (62/103) of primary MBs compared with normal study looking at 201 primary human MBs, in which cerebellum, whereas SFRP1 promoter region methylation losses targeting chromosome 8, including the SFRP1

Oncogene SFRP gene silencing in medulloblastoma PN Kongkham et al 3022

Oncogene SFRP gene silencing in medulloblastoma PN Kongkham et al 3023 locus, were observed in 31/201 (15%) tumors (Figure 3b) tumor suppressor genes in MB and verified their (Northcott et al., 2009b). Chromosome 8p loss occurred methylation in MB cell lines and primary tumors. with increased frequency in Group C and Group D Functionally, SFRP1, SFRP2 and SFRP3 expression tumors, and was associated with a trend toward reduced limited WNT signaling, reduced MB cell proliferation SFRP1 expression (Figure 3c). SFRP1 may also be and anchorage-independent growth in vitro. SFRP1 influenced by transcription factor-mediated repression. expression reduced tumor burden leading to increased The MYC transcription factor is known to bind to the overall survival in vivo. We observed low-level expres- SFRP1 promoter region, strongly repressing SFRP1 sion of SFRP1 in 60% of primary human MB samples transcription (Cowling et al., 2007). On the basis of in comparison with normal cerebellum. Reduced SFRP1 Exon Array data, the top quartile of MYC expressors expression was observed in the WNT, Group C and (‘high MYC group’) showed reduced SFRP1 expression Group D MB molecular subgroups in comparison with compared with the bottom quartile of MYC expressors SHH-driven tumors. We acknowledge the limitations of (‘low MYC group’) (Figure 3d). Although not examined using nonmatched human fetal and adult whole specifically here, modifications to the histone code may cerebellar tissue samples for normal controls as one also contribute to SFRP1 gene silencing in the absence area of weakness in this and other studies in this field. of promoter region methylation. Given the difficulties in obtaining normal cerebellar tissue samples as well as an incomplete understanding SFRP1 expression impairs MB tumor formation in vivo regarding the cell of origin for various MB subtypes, To examine the effect of SFRP1 expression on MB these samples represent one of the best normal human growth in vivo, we employed both flank and orthotopic tissue controls currently available. intracerebellar xenograft models (Figure 4). ONS76 cells In addition to epigenetic mechanisms including stably transfected with empty vector or pcDNA3.1- promoter region methylation, SFRP1 expression may SFRP1-FLAG were injected into the left and right hind be influenced by gene deletion events or transcription flanks of immunodeficient mice, respectively. As seen in factor-mediated silencing. Aberrant canonical WNT Figure 4a, 5/5 tumors formed in the left flank when signaling has been firmly implicated in MB pathogen- injected with empty vector-transfected cells. In compar- esis. Our study is the first to identify loss of pathway ison, SFRP1 re-expressing cells formed tumors in only inhibition due to SFRP gene silencing as an additional 1/5 mice. Mice engrafted intracranially with empty mechanism contributing to dysregulation of WNT vector-transfected D283 MB cells quickly succumbed signaling in MB. to symptoms related to tumor progression, with a mean overall survival of 27.6±1.3 days. SFRP1 expression in D283 cells conferred a significant survival advan- Conflict of interest tage in this orthotopic xenograft model (Po0.0001) (Figure 4b). All mice engrafted with SFRP1-expressing The authors declare no conflict of interest. D283 cells remained alive and asymptomatic for 60 days, at which time the data were censored and specimens collected for further characterization. All mice engrafted with empty vector-transfected cells Acknowledgements developed large cerebellar lesions (Figure 4c). In contrast, tumors from SFRP1-transfected cells were This work was supported by grants from the Canadian Cancer considerably smaller and in some cases not grossly Society, National Cancer Institute of Canada (019073), the visible on hematoxylin and eosin staining (Figure 4c). Pediatric Brain Tumor Foundation, the Wiley Fund at the Hospital for Sick Children and B.r.a.i.n.child. James Rutka is a scientist of the Canadian Institutes of Health Research. Paul Kongkham was supported by the Surgeon Scientist Program Conclusions (University of Toronto), the National Cancer Institute of Canada Terry Fox Foundation clinical research fellowship and In summary, using a genome-wide epigenetic strategy, a research studentship from RESTRACOMP, The Hospital we identified SFRP1, SFRP2 and SFRP3 as candidate for Sick Children, Toronto.

Figure 4 (a) ONS76 flank xenografts showing tumor formation in 5/5 mice in the left flank injected with ONS76 stably transfected with empty vector (arrow), compared with 1/5 mice in the right flank injected with ONS76 stably expressing SFRP1 (arrowhead). In total, 2 Â 106 ONS76 cells stably transfected with empty vector or pcDNA3.1-SFRP1-FLAG were injected into the flank of nude mice and followed up for 10 weeks. (b) Kaplan–Meier survival curve for intracerebellar xenografts using stably transfected D283-empty vector cells (EV) or D283-SFRP1 cells (SFRP1). A significant survival advantage was conferred by SFRP1 expression, with all SFRP1 mice remaining asymptomatic at 60 days (Po0.0001). 1 Â 105 D283 cells stably transfected with empty vector or pcDNA3.1-SFRP1- FLAG were stereotactically injected orthotopically into the midline cerebella of 10 nude mice per group. Mice were killed when displaying evidence of significant intracranial mass effect or a loss of 20% body weight was observed. The experiment was censored at 2 months, when remaining mice were killed and examined by hematoxylin and eosin staining and immunostaining for SFRP1 (R&D Systems, Minneapolis, MN, USA). (c) Hematoxylin and eosin staining and immunohistochemistry for SFRP1 in representative intracerebellar xenograft specimens. Empty vector controls produced large tumors (arrow) with leptomeningeal spread (arrowhead). SFRP1 expression significantly reduced tumor burden (arrow).

Oncogene SFRP gene silencing in medulloblastoma PN Kongkham et al 3024 References

BaezaN,MasuokaJ,KleihuesP,OhgakiH.(2003).AXIN1mutations result in TCF-dependent transcription in medulloblastomas. Int J but not deletions in cerebellar medulloblastomas. Oncogene 22: 632–636. Cancer 121: 284–291. Bovolenta P, Esteve P, Ruiz JM, Cisneros E, Lopez-Rios J. (2008). Kongkham PN, Northcott PA, Ra YS, Nakahara Y, Mainprize TG, Beyond Wnt inhibition: new functions of secreted Frizzled-related Croul SE et al. (2008). An epigenetic genome-wide screen identifies in development and disease. J Cell Sci 121: 737–746. SPINT2 as a novel tumor suppressor gene in pediatric medullo- Chang Q, Pang JC, Li KK, Poon WS, Zhou L, Ng HK. (2005). blastoma. Cancer Res 68: 9945–9953. Promoter hypermethylation profile of RASSF1A, FHIT, and Kool M, Koster J, Bunt J, Hasselt NE, Lakeman A, van Sluis P et al. sFRP1 in intracranial primitive neuroectodermal tumors. Hum (2008). Integrated genomics identifies five medulloblastoma sub- Pathol 36: 1265–1272. types with distinct genetic profiles, pathway signatures and Cowling VH, D’Cruz CM, Chodosh LA, Cole MD. (2007). c-Myc clinicopathological features. PLoS ONE 3: e3088. transforms human mammary epithelial cells through repression of Lee Y, Miller HL, Jensen P, Hernan R, Connelly M, Wetmore C et al. the Wnt inhibitors DKK1 and SFRP1. Mol Cell Biol 27: 5135–5146. (2003). A molecular fingerprint for medulloblastoma. Cancer Res Dahl E, Wiesmann F, Woenckhaus M, Stoehr R, Wild PJ, Veeck J 63: 5428–5437. et al. (2007). Frequent loss of SFRP1 expression in multiple human McMahon AP, Bradley A. (1990). The Wnt-1 (int-1) proto-oncogene is solid tumours: association with aberrant promoter methylation in required for development of a large region of the mouse brain. Cell renal cell carcinoma. Oncogene 26: 5680–5691. 62: 1073–1085. Eberhart CG, Tihan T, Burger PC. (2000). Nuclear localization and Northcott PA, Fernandez LA, Hagan JP, Ellison DW, Grajkowska W, mutation of beta-catenin in medulloblastomas. J Neuropathol Exp Gillespie Y et al. (2009a). The miR-17/92 polycistron is up-regulated Neurol 59: 333–337. in sonic hedgehog-driven medulloblastomas and induced by N-myc Ellison DW, Onilude OE, Lindsey JC, Lusher ME, Weston CL, Taylor in sonic hedgehog-treated cerebellar neural precursors. Cancer Res RE et al. (2005). beta-Catenin status predicts a favorable outcome in 69: 3249–3255. childhood medulloblastoma: the United Kingdom Children’s Cancer Northcott PA, Nakahara Y, Wu X, Feuk L, Ellison DW, Croul S et al. Study Group Brain Tumour Committee. JClinOncol23: 7951–7957. (2009b). Multiple recurrent genetic events converge on control Fan X, Eberhart CG. (2008). Medulloblastoma stem cells. J Clin Oncol of histone lysine methylation in medulloblastoma. Nat Genet 41: 26: 2821–2827. 465–472. Gilbertson RJ, Ellison DW. (2008). The origins of medulloblastoma Suzuki H, Watkins DN, Jair KW, Schuebel KE, Markowitz SD, Chen subtypes. Annu Rev Pathol 3: 341–365. WD et al. (2004). Epigenetic inactivation of SFRP genes Habas R, Dawid IB. (2005). Dishevelled and Wnt signaling: is the allows constitutive WNT signaling in colorectal cancer. Nat Genet nucleus the final frontier? J Biol 4:2. 36: 417–422. Hamilton SR, Liu B, Parsons RE, Papadopoulos N, Jen J, Powell SM Thompson MC, Fuller C, Hogg TL, Dalton J, Finkelstein D, Lau CC et al. (1995). The molecular basis of Turcot’s syndrome. N Engl J et al. (2006). Genomics identifies medulloblastoma subgroups Med 332: 839–847. that are enriched for specific genetic alterations. J Clin Oncol 24: Katoh Y, Katoh M. (2006). WNT antagonist, SFRP1, is Hedgehog 1924–1931. signaling target. Int J Mol Med 17: 171–175. Vibhakar R, Foltz G, Yoon JG, Field L, Lee H, Ryu GY et al. (2007). Koch A, Hrychyk A, Hartmann W, Waha A, Mikeska T, Schuller U Dickkopf-1 is an epigenetically silenced candidate tumor suppressor et al. (2007). Mutations of the Wnt antagonist AXIN2 (Conductin) gene in medulloblastoma. Neuro Oncol 9: 135–144.

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

Oncogene