Oncogene (2008) 27, 4745–4751 & 2008 Macmillan Publishers Limited All rights reserved 0950-9232/08 $30.00 www.nature.com/onc SHORT COMMUNICATION High-resolution, dual-platform aCGH analysis reveals frequent HIPK2 amplification and increased expression in pilocytic astrocytomas

H Deshmukh1,5, TH Yeh2,4,5,JYu1,5, MK Sharma1, A Perry1, JR Leonard3, MA Watson1, DH Gutmann2 and R Nagarajan1

1Department of Pathology and Immunology, Washington University School of Medicine, St Louis, MO, USA; 2Department of Neurology, Washington University School of Medicine, St Louis, MO, USA; 3Department of Neurosurgery, Washington University School of Medicine, St Louis, MO, USA and 4Department of Neurology, Chang Gung Memorial Hospital and University, Taipei, Taiwan

Pilocytic astrocytomas (PAs, WHO grade I) are the most Astrocytic tumors (gliomas) are the most common common brain tumors in the pediatric and adolescent primary glial neoplasm affecting children and young population, accounting for approximately one-fifth of adults, accounting for 16–21% of central nervous central nervous system tumors. Because few consistent system tumors in individuals less than 19 years of age. molecular alterations have been identified in PAs com- These low-grade tumors exhibit rare mitotic figures, but pared to higher grade gliomas, we performed array often have microvascular proliferation and microglial comparative genomic hybridization using two independent infiltration. In contrast to other astrocytic neoplasms, commercial array platforms. Although whole chromo- pilocytic astrocytomas (PAs) rarely progress to more somal gains and losses were not observed, a 1-Mb malignantgliomas and have even been reportedto amplified region of 7q34 was detected in multiple patient regress without treatment when they occur in patients samples using both array platforms. Copy-number gain with neurofibromatosis-1. This characteristic behavior was confirmed in an independent tumor sample set by implies that PA may be a molecularly distinct glial quantitative PCR, and this amplification was correlated to neoplasm. both increased mRNA and protein expression of HIPK2, The unique clinical behavior of PAs is also reflected in a -interacting associated with the infrequent identification of genomic/genetic changes malignancy, contained within this . Furthermore, in these tumors. Although TP53 and PTEN overexpression of wild-type HIPK2, but not a kinase- are common in higher grade gliomas (Cheng et al., 2000; inactive mutant, in a glioma cell line conferred a growth Tada et al., 2003; Arjona et al., 2006), these are advantage in vitro. Collectively, these results illustrate the rarely altered in PA (Ishii et al., 1998; Hayes et al., 1999; power and necessity of implementing high-resolution, Cheng et al., 2000; von Deimling et al., 2000; Wemmert multiple-platform genomic analyses to discover small et al., 2006). Similarly, although chromosomal aberra- and subtle, but functionally significant, genomic altera- tions or amplification/deletion of specific genes related tions associated with low-grade tumor formation and to cell cycle control, growth factor receptors and signal growth. transduction pathways are commonly identified in other Oncogene (2008) 27, 4745–4751; doi:10.1038/onc.2008.110; glioma subtypes (Koschny et al., 2002; Schmidt et al., published online 14 April 2008 2002; Ohgaki et al., 2004; Ohgaki, 2005), most cytogenetic studies have shown that the majority of Keywords: pilocytic astrocytomas; aCGH; HIPK2 PAs have normal karyotypes. Structural aberrations amplification involving 6, 7, 8, 11, 17 and 19 have been reported in a few cases (Jenkins et al., 1989; Thiel et al., 1992; Agamanolis and Malone, 1995), and the most consistently identified cytogenetic changes involve gains on chromosomes 7 and 8 (White et al., 1995; Zattara-Cannoni et al., 1998; Sanoudou et al., 2000; Correspondence: Dr DH Gutmann, Department of Neurology, Jones et al., 2006). We have previously identified a Washington University School of Medicine, 660 S. Euclid Avenue, transcript on 8 whose expression is 306 Biotechnology Building, Campus Box 8111, St Louis, MO 63110, increased in PA (Sharma et al., 2006); however, no USA. E-mail: [email protected] or Dr R Nagarajan, Department specific genes have been implicated on . of Pathology and Immunology, Washington University School of In an effort to identify recurrent PA-associated Medicine, 660 S. Euclid Avenue, Campus Box 8118, StLouis, MO genetic abnormalities in a more comprehensive manner, 63110, USA. we employed high-resolution, array comparative geno- E-mail: [email protected] 5These authors contributed equally to this work. mic hybridization (aCGH) using two different array Received 11 December 2007; revised 24 January 2008; accepted 1 March platforms from NimbleGen and Affymetrix at a 2008; published online 14 April 2008 resolution of approximately 6 and 2.5 kb, respectively. HIPK2 amplification in pilocytic astrocytomas H Deshmukh et al 4746 Ten patient-matched cerebellar PAs and germline DNA platforms (Figure 2b). Importantly, this finding was sample pairs from patients with age less than 18 years confirmed in the independent set where 42.6% (26/61) of were selected for this study. These tumors were these tumors had HIPK2 amplification (mean ¼ 2.6-fold specifically chosen based on their common histologic amplification) (Figure 3a). Quantitative RT–PCR and appearance and similar brain location and patient age. co-analysis of copy number and microarray expression All samples were subject to simultaneous aCGH using a data further confirmed that HIPK2 amplification 385K high-resolution NimbleGen oligonucleotide array correlated with increased HIPK2 mRNA levels in the and the Affymetrix Mapping 500K SNP Array Set. To same tumor (Supplementary Figure 1, Supplementary visualize large, contiguous amplifications and deletions Data). across the genome, individual probe-level copy-number To determine whether the increase in HIPK2 copy values of the tumors were plotted after internally number and RNA expression was observed also atthe normalizing each tumor to its germline control protein level, immunohistochemistry on brain tumor (Figure 1). Although most samples did not have specimens was performed using three separate tissue chromosome-wide copy number gains or losses, a single microarrays. First, using a tissue microarray containing sample (patient ID 963) had a copy number gain of normal brain and PAs, 41/63 (65%) of PA specimens many single-nucleotide polymorphism (SNP) markers exhibited increased HIPK2 expression relative to normal on chromosomes 6 and 21 that was confirmed by brain. Positive HIPK2 immunostaining was defined as segmentation analysis of the NimbleGen data. Thus, no greater than 25% of the tumor cells with nuclear HIPK2 recurrent, large-scale amplifications or deletions were expression. Eighty percentage of the PAs that scored observed consistently on both array platforms in the PA positive by this criteria actually contained greater than sample set, a result previously confirmed by lower 75% tumor cells with nuclear HIPK2 expression resolution cytogenetic-based studies. (Figure 3b panels i–iv). These results were highly Although multiple focal abnormalities in individual reproducible in that we observed a 90% concordance in cases were observed on either platform (Figure 1), it was immunostaining among multiple cores from the same difficulttodiscern regions which were recurrently tumor represented on the tissue microarray. amplified or deleted in PAs consistently across both In contrast to the high level of HIPK2 expression technologies based on visual inspection alone. Thus, copy observed in PAs (WHO grade I glioma), significantly number results from both dChip and CNAT algorithms lower levels of HIPK2 immunopositivity were observed were first combined to identify SNP markers that were in other glioma subtypes and histological grades. In this called amplified or deleted in more than one sample. regard, 23 and 27% of the grade III (anaplastic A total of 118 SNP markers on chromosome 7 and 23 astrocytoma) and IV (glioblastoma multiforme) glio- markers on chromosomes 1, 7, 12, 15 and 21 were mas, respectively, and 12% of oligodendrogliomas recurrently amplified or deleted, respectively, by both exhibited HIPK2 overexpression relative to non-malig- tools (Supplementary Table 1, Supplementary Data). nantbrain tissue. We observed no HIPK2 overexpres- Next, to identify bona fide regions of recurrent amplifica- sion in ependymomas. These results demonstrate that tion and deletion, segment-level results from each HIPK2 is overexpressed in a large proportion of PAs corresponding NimbleGen array were co-analyzed with compared to other glial tumors; however, HIPK2 segment-level results from dChip, CNAT and CNAG overexpression is nota specific signatureof PAs. (Figure 2a). Although no deletions were observed by To determine whether HIPK2 overexpression confers a segment-level analysis, a single region on 7q34 (chromo- growth advantage in vitro, we performed clonogenic some 7: 138 151 200–139 456 000, NCBI build 35) was assays using U87 human glioma cells, which demonstrate amplified in 8 out of 10 patient-matched tumor-germline a low level of endogenous HIPK2 expression (data not samples (Figure 2a, ‘SNP Segments’). This same region shown). Equal numbers of U87 cells were transfected with was found amplified in 6 outof 10 samples by NimbleGen either empty vector, active HIPK2 or a kinase-inactivated aCGH (Figure 2a, ‘aCGH’). The mostfrequently HIPK2 mutant (K221R) (D’Orazi et al., 2002) along with amplified SNP and aCGH segments were localized to a plasmid containing antibiotic resistance. Consistent with the upstream region of the homeodomain-interacting the prediction that HIPK2 promotes cell growth, we protein kinase 2 (HIPK2)gene. observed a 1.8-fold increase (mean of three independent To validate HIPK2 amplification in PA, quantitative experiments) in the number of colonies following HIPK2 PCR (qPCR) was performed using matched peripheral overexpression (Figure 4). There was no effectof blood cell and tumor from the original PA kinase-inactivated HIPK2 mutant or empty vector sample setas well as from an independentsetof 61 PAs. overexpression on U87 colony formation. The cutoff for HIPK2 gain/loss in the tumor samples The observation that HIPK2 overexpression leads to was setusing themean HIPK2 copy number (2.03)±3 increased cell growth in a glioma cell line is interesting in standard deviations (0.11) from the 10 peripheral blood light of the reported roles of HIPK2 in growth cell samples (Konigshoff et al., 2003). Using these regulation. HIPK2 is a / kinase that criteria (that is HIPK2 > ¼ 2.37 to be called a gain or functions as a transcriptional regulator by multiple HIPK2 o ¼ 1.68 to be called a loss), 60% (6/10) of mechanisms. First, it acts as a co-repressor by directly tumors in the original PA sample set had a HIPK2 copy interacting with transcription factors such as Brn3a number gain, and this level of amplification was (Wiggins et al., 2004) and c-Ski (Harada et al., 2003). comparable to that observed in both array-based Second, it mediates transcriptional activation or

Oncogene Deshmukh H HIPK2 mlfiaini ioyi astrocytomas pilocytic in amplification tal et

908 791 1019 963 1125 13060 1163 4118 719 1831 908 791 1019 963 1125 13060 1163 4118 719 1831

Figure 1 DNAs from 10 patient-matched PAs and peripheral blood leukocytes were extracted, labeled and hybridized to NimbleGen ‘2006-02-14_HG17_WG_CGH’ arrays ( build 17, National Center for Biotechnology Information build 35) and to Affymetrix GeneChip Human Mapping 500K array sets. After scanning and quantification of signal intensities were performed, tumor copy number was plotted using NimbleGen (unaveraged, Z-score normalized probe level data (a)) and Affymetrix (dChip-derived copy number data (b)) data. The magnitude of amplifications and deletions in the tumor genome is color-coded using a gradient from green or red, respectively, to white. CGH, comparative genomic hybridization; PA, pilocytic astrocytoma. Oncogene 4747 HIPK2 amplification in pilocytic astrocytomas H Deshmukh et al 4748 a 5' 137992226

3' 140067396

b NimbleGen array CGH array NimbleGen Affymetrix SNP arrays Affymetrix qPCR copy number qPCR copy Tumor location Patient ID Gender Age

719 3 F Cerebellum 2.01 2.24 2.62 791 13 M Cerebellum 2.31 2.74 2.98 908 12 F Cerebellum 1.99 1.93 1.81 963 12 M Cerebellum 2.28 2.23 2.28 1019 16 M Cerebellum 2.00 2.08 1.79 1125 9 F Cerebellum 2.30 2.57 1.87 1163 8 F Cerebellum 2.27 2.34 2.40 1831 11 M Cerebellum 2.31 2.40 2.54 4118 9 F Cerebellum 2.33 2.50 2.40 13060 9 M Cerebellum 2.37 2.48 2.46

Figure 2 (a) Detailed view of the 7q34 locus amplified in multiple samples. SNP markers and segments identified using the Affymetrix and NimbleGen array data are shown from left to right. Amplified SNP markers are shown in black whereas non-amplified SNP markers are shown in gray. Genes in the region are displayed on the far right. CGH-segMNT was used to identify amplified and deleted segments using the NimbleGen data. For Affymetrix data, copy number analysis was performed using dChip (Lin et al., 2004), CNAG (Nannya et al., 2005) and the Copy Number Analysis Tool (CNAT). To facilitate cross-platform comparison, probes from both array platforms were mapped to NCBI Build 35, and SNP marker data were converted to segment data using DNAcopy (Venkatraman and Olshen, 2007). Recurrent, minimally overlapping regions were identified by analyzing segments identified using either array platform. (b) Patient characteristics (age, gender and tumor location) and HIPK2 copy number levels from Affymetrix, NimbleGen and DNA qPCR are shown. HIPK2 is consistently amplified across all three detection platforms in 5/10 samples (Copy number 42.25 in Affymetrix, 42.31 in NimbleGen and >2.37 in qPCR are in bold according to cutoffs based on the reference gene DOCK10). CGH, comparative genomic hybridization; qPCR, quantitative PCR; SNP, single-nucleotide polymorphism.

Oncogene HIPK2 amplification in pilocytic astrocytomas H Deshmukh et al 4749

Figure 3 (a) DNA qPCR-based HIPK2 copy number data are plotted for matched peripheral blood cells (10-PBC) and tumor DNAs (10-PA) of the 10 PA patients and 61 independent PAs (61-PA). The cutoff for DNA copy number was set using the mean±3 s.d. of the copy numbers derived from the 10 PBC samples. A 20-ml reaction mixture for qPCR was composed of 10 mlof2Â TaqMan universal PCR master mix (Applied Biosystems Inc.), 10 ml of primer and probe mix (600 nM each forward and reverse primers, 200 nM specific TaqMan probe at final concentration) and 10 ng of DNA. All real-time PCR assays were performed in triplicate on an ABI PRISM 7900HT Sequence Detector System (ABI) as follows: 50 1C for 2 min, 95 1C for 10 min and 40 cycles of both 15 s at 95 1C and 60 sec at60 1C. The reference gene, DOCK10, which was chosen to normalize the DNA input between individual DNA samples, is on chromosome 2 where no significant gains/losses were observed with both NimbleGen and Affymetrix array data in our 10 PA cases and in another study (Jones et al., 2006). (b) The upper panel demonstrates representative PAs with different percentages of cells within the tumor demonstrating HIPK2 nuclear staining; (i) >75%; (ii) 50–75%; (iii) 25–50%; (iv) o25% (scale, 100 mm). The lower panel lists the number and percentage of tumors with positive staining for HIPK2 where positive nuclear staining was observed in >25% of the cells in the tumor specimen. Higher rates of positive nuclear staining for HIPK2 were observed in PAs than in higher grade astrocytomas or other tumors of neurological origin. HIPK2 immunohistochemistry was performed on glioma tissue microarrays after antigen retrieval, and slides were hematoxylin-counterstained prior to determining positive staining (See Supplementary Methods for further details). PA, pilocytic astrocytoma; qPCR, quantitative PCR. repression through of transcription induces or represses , inhibits cell proliferation, factors such as , Pax6, c-Myb and AML1 (Kim regulates segmental identity and is crucial for et al., 1998; Calzado et al., 2007). As a result, HIPK2 hematopoiesis.

Oncogene HIPK2 amplification in pilocytic astrocytomas H Deshmukh et al 4750

Figure 4 (a) Clonogenic assays were used to demonstrate that HIPK2 expression increases cell growth. U87 glioblastoma cells (1 Â 105) per well were co-transfected with 2 mg of equimolar amounts of pCMV-HIPK2, pCMV-HIPK2m (K221R) or empty vector in combination with pBABE-puro (Addgene, Cambridge, MA, USA) using the Fugene 6 reagent (Roche Diagnostics, Indianapolis, IN, USA) before selection with puromycin (0.7 mg/ml; Sigma-Aldrich, StLouis, MO, USA) for 14–21 days. Surviving colonies were stained with crystal violet and counted. Representative plates of HIPK2, HIPK2m or empty vector transfections are shown. (b) Average clone counts across three independent experiments are plotted. Statistical significance among different groups was analyzed by one-way ANOVA followed by Bonferroni’s test. A P-value o0.05 was considered significant. Statistically significant differences were noted between HIPK2-transfected cells and either vector or HIPK2m-transfected cells.

Functional studies have demonstrated that HIPK2 50-fold amplification of genes observed in other tumors functions as a tumor suppressor in some cell types (Albertson, 2006), the low, but functionally significant (Pierantoni et al., 2002; Wei et al., 2007). However, in amplification of HIPK2 in PA (average copy the central nervous system, Hipk2 loss leads to selective number ¼ 2.57 by qPCR) would likely be missed or loss of midbrain dopamine neurons as a resultof discounted as ‘noise’ in studies that utilize lower increased apoptosis mediated by the inability of the resolution array designs or a single-array platform. receptor-activated Smads (for example, Smad2 and Although the low level of HIPK2 amplification observed Smad3) to regulate transcription in response to TGFb3 may reflectan averaged value of inherenttumorgenome (Zhang et al., 2007). Consistent with our results, HIPK2 heterogeneity, it is also possible that such small overexpression leads to increased glioma cell growth quantitative copy number increases of regulatory genes, in vitro. Further studies will be required to define the like HIPK2, may result in significant perturbations of mechanism underlying HIPK2-positive growth regula- downstream targets that contribute to the pathophysio- tion in astrocytes and glioma cells. logy of PAs. The paucity of genomic alterations that This study leverages significant advances both in were found in our sample set may indicate that only a experimental design and in technology platform to few genes drive the tumorigenic process of PA or that identify subtle genomic imbalances that may occur in other mechanisms such as gene methylation or sequence low-grade malignancies such as PAs. First, the use of may be more prevalent in these tumors. In patient-matched germline and tumor DNA permitted us summary, our study highlights the need to employ to clearly distinguish somatic amplifications and dele- comprehensive and complementary technologies to tions from a multitude of possible germline copy identify subtle copy number changes that underlie the number variations that would be detected using an molecular pathogenesis of low-grade tumors. unrelated ‘normal’ reference pool (Sebat et al., 2004). Second, the utilization of two independent, high- Acknowledgements resolution genome-wide array platforms enabled the identification of small, subtle alterations with increased HIPK2 plasmids were generously provided by Dr Gabriella confidence. In striking contrast to patterns of 10- to D’Orazi (Regina Elena Cancer Institute, Rome Italy).

Oncogene HIPK2 amplification in pilocytic astrocytomas H Deshmukh et al 4751 References

Agamanolis DP, Malone JM. (1995). Chromosomal abnormalities Ohgaki H, Dessen P, Jourde B, Horstmann S, Nishikawa T, Di Patre PL in 47 pediatric brain tumors. Cancer Genet Cytogenet 81: 125–134. et al. (2004). Genetic pathways to glioblastoma: a population-based Albertson DG. (2006). Gene amplification in cancer. Trends Genet 22: study. Cancer Res 64: 6892–6899. 447–455. Pierantoni GM, Bulfone A, Pentimalli F, Fedele M, Iuliano R, Arjona D, Rey JA, Taylor SM. (2006). Early genetic changes involved in Santoro M et al. (2002). The homeodomain-interacting protein low-grade astrocytic tumor development. Curr Mol Med 6: 645–650. kinase 2 gene is expressed late in embryogenesis and preferentially in Calzado MA, Renner F, Roscic A, Schmitz ML. (2007). HIPK2: retina, muscle, and neural tissues. Biochem Biophys Res Commun a versatile switchboard regulating the transcription machinery and 290: 942–947. cell death. Cell Cycle 6: 139–143. Sanoudou D, Tingby O, Ferguson-Smith MA, Collins VP, Cheng Y, Pang JC, Ng HK, Ding M, Zhang SF, Zheng J et al. (2000). Coleman N. (2000). Analysis of pilocytic astrocytoma by compara- Pilocytic astrocytomas do not show most of the genetic changes tive genomic hybridization. Br J Cancer 82: 1218–1222. commonly seen in diffuse astrocytomas. Histopathology 37: 437–444. SchmidtMC, Antweiler S, Urban N, Mueller W, Kuklik A, D’Orazi G, Cecchinelli B, Bruno T, Manni I, Higashimoto Y, Saito S Meyer-Puttlitz B et al. (2002). Impactof genotype and morphology on et al. (2002). Homeodomain-interacting protein kinase-2 phosphory- the prognosis of glioblastoma. J Neuropathol Exp Neurol 61: 321–328. lates p53 at Ser 46 and mediates apoptosis. Nat Cell Biol 4: 11–19. SebatJ, Lakshmi B, Troge J, Alexander J, Young J, Lundin P et al. Harada J, Kokura K, Kanei-Ishii C, Nomura T, Khan MM, Kim Y (2004). Large-scale copy number polymorphism in the human et al. (2003). Requirementof theco-repressor homeodomain- genome. Science 305: 525–528. interacting protein kinase 2 for ski-mediated inhibition of bone Sharma MK, Watson MA, Lyman M, Perry A, Aldape KD, Deak F morphogenetic protein-induced transcriptional activation. J Biol et al. (2006). Matrilin-2 expression distinguishes clinically relevant Chem 278: 38998–39005. subsets of pilocytic astrocytoma. Neurology 66: 127–130. Hayes VM, Dirven CM, Dam A, Verlind E, Molenaar WM, Mooij JJ Tada K, Kochi M, Saya H, Kuratsu J, Shiraishi S, Kamiryo T et al. et al. (1999). High frequency of TP53 mutations in juvenile pilocytic (2003). Preliminary observations on genetic alterations in pilocytic astrocytomas indicates role of TP53 in the development of these astrocytomas associated with neurofibromatosis 1. Neuro Oncol 5: tumors. Brain Pathol 9: 463–467. 228–234. Ishii N, Sawamura Y, Tada M, Daub DM, Janzer RC, Thiel G, Losanowa T, Kintzel D, Nisch G, Martin H, Vorpahl K et al. Meagher-Villemure M et al. (1998). Absence of p53 gene mutations (1992). Karyotypes in 90 human gliomas. Cancer Genet Cytogenet in a tumor panel representative of pilocytic astrocytoma diversity 58: 109–120. using a p53 functional assay. Int J Cancer 76: 797–800. Venkatraman ES, Olshen AB. (2007). A faster circular binary Jenkins RB, Kimmel DW, Moertel CA, Schultz CG, Scheithauer BW, segmentation algorithm for the analysis of array CGH data. Kelly PJ et al. (1989). A cytogenetic study of 53 human gliomas. Bioinformatics 23: 657–663. Cancer Genet Cytogenet 39: 253–279. von Deimling A, Fimmers R, SchmidtMC, Bender B, Fassbender F, Jones DT, Ichimura K, Liu L, Pearson DM, PlantK, Collins VP. Nagel J et al. (2000). Comprehensive allelotype and genetic anaysis (2006). Genomic analysis of pilocytic astrocytomas at 0.97 Mb of 466 human nervous system tumors. J Neuropathol Exp Neurol 59: resolution shows an increasing tendency toward chromosomal copy 544–558. number change with age. J Neuropathol Exp Neurol 65: 1049–1058. Wei G, Ku S, Ma GK, Saito S, Tang AA, Zhang J et al. (2007). HIPK2 Kim YH, Choi CY, Lee SJ, Conti MA, Kim Y. (1998). represses beta-catenin-mediated transcription, epidermal stem cell Homeodomain-interacting protein kinases, a novel family of expansion, and skin tumorigenesis. Proc Natl Acad Sci USA 104: co-repressors for homeodomain transcription factors. J Biol Chem 13040–13045. 273: 25875–25879. WemmertS, Romeike BF, Ketter R, Steudel WI, Zang KD, UrbschatS. Konigshoff M, Wilhelm J, Bohle RM, Pingoud A, Hahn M. (2003). (2006). Intratumoral genetic heterogeneity in pilocytic astrocytomas HER-2/neu gene copy number quantified by real-time PCR: revealed by CGH-analysis of microdissected tumor cells and FISH on comparison of gene amplification, heterozygosity, and immuno- tumor tissue sections. Int J Oncol 28: 353–360. histochemical status in breast cancer tissue. Clin Chem 49: 219–229. White FV, Anthony DC, Yunis EJ, Tarbell NJ, Scott RM, Koschny R, Koschny T, Froster UG, Krupp W, Zuber MA. (2002). Schofield DE. (1995). Nonrandom chromosomal gains in pilocytic Comparative genomic hybridization in glioma: a meta-analysis of astrocytomas of childhood. Hum Pathol 26: 979–986. 509 cases. Cancer Genet Cytogenet 135: 147–159. Wiggins AK, Wei G, Doxakis E, Wong C, Tang AA, Zang K et al. Lin M, Wei LJ, Sellers WR, Lieberfarb M, Wong WH, Li C. (2004). (2004). Interaction of Brn3a and HIPK2 mediates transcriptional dChipSNP: significance curve and clustering of SNP-array-based repression of sensory neuron survival. J Cell Biol 167: 257–267. loss-of-heterozygosity data. Bioinformatics 20: 1233–1240. Zattara-Cannoni H, Gambarelli D, Lena G, Dufour H, Choux M, Nannya Y, Sanada M, Nakazaki K, Hosoya N, Wang L, Hangaishi A Grisoli F et al. (1998). Are juvenile pilocytic astrocytomas benign et al. (2005). A robust algorithm for copy number detection using tumors? A cytogenetic study in 24 cases. Cancer Genet Cytogenet high-density oligonucleotide single nucleotide polymorphism geno- 104: 157–160. typing arrays. Cancer Res 65: 6071–6079. Zhang J, Pho V, Bonasera SJ, Holtzman J, Tang AT, Hellmuth J et al. Ohgaki H. (2005). Genetic pathways to glioblastomas. Neuropathology (2007). Essential function of HIPK2 in TGFbeta-dependent survival 25: 1–7. of midbrain dopamine neurons. Nat Neurosci 10: 77–86.

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

Oncogene