Homozygously deleted gene DACH1 regulates tumor-initiating activity of glioma cells

Abstract

Loss or reduction in function of tumor suppressor genes contributes to tumorigenesis. Here, by allelic DNA copy number analysis using single-nucleotide polymorphism genotyping array and mass spectrometry, we report homozygous deletion in glioblastoma multiformes at chromosome 13q21, where DACH1 gene is located. We found decreased cell proliferation of a series of glioma cell lines by forced expression of DACH1. We then generated U87TR-Da glioma cells, where DACH1 expression could be activated by exposure of the cells to doxycycline. Both ex vivo cellular proliferation and in vivo growth of s.c. transplanted tumors in mice are reduced in U87TR-Da cells with DACH1 expression (U87-DACH1-high), compared with DACH1-nonexpressing U87TR-Da cells (U87-DACH1-low). U87-DACH1-low cells form spheroids with CD133 and Nestin expression in serum-free medium but U87-DACH1-high cells do not. Compared with spheroid-forming U87-DACH1-low cells, adherent U87-DACH1-high cells display lower tumorigenicity, indicating DACH1 decreases the number of tumor-initiating cells. Gene expression analysis and chromatin immunoprecipitation assay reveal that fibroblast growth factor 2 (FGF2/bFGF) is transcriptionally repressed by DACH1, especially in cells cultured in serum-free medium. Exogenous bFGF rescues spheroid-forming activity and tumorigenicity of the U87-DACH1-high cells, suggesting that loss of DACH1 increases the number of tumor-initiating cells through transcriptional activation of bFGF. These results illustrate that DACH1 is a distinctive tumor suppressor, which does not only suppress growth of tumor cells but also regulates bFGF-mediated tumor-initiating activity of glioma cells.

Glioblastoma multiformes (GBMs), the most frequent primary malignant brain tumor in adults, are aggressive and highly invasive tumors (1). Genetic alterations of GBMs, including aberration of DNA copy number such as gene amplifications, loss of heterozygosity (LOH), and homozygous deletions, leads to activation of oncogenes and inactivation of tumor suppressor genes (1–3). DNA copy number analysis by single-nucleotide polymorphism (SNP) genotyping enables the high-resolution analysis of allelic DNA copy number and has been used to obtain a genome-wide view of DNA copy number alterations in human cancers (4–10). Particularly, pairwise analysis of normal and tumor DNAs is crucial in detecting homozygous deletion in clinical specimens, because infiltrating nontumorous cells are significant in GBMs.

In this study we examined the allelic copy number of paired glioma and blood DNAs by SNP genotyping array analysis by using Genome Imbalanced Map (GIM) algorithm (5, 11), which could calculate the signal ratio of SNP genotyping array in an allelic manner. We identified a unique homozygous deletion at DACH1 gene region on chromosome 13q21, and we demonstrated forced expression of DACH1 reduced proliferation of cultured glioma cells and in vivo tumor growth in ortholotopic xenograft model. We also found that DACH1 inhibited formation of tumor-initiating spheroids, presumably by directly repressing expression of fibroblast growth factor-2 (FGF2), suggesting DACH1 is a unique tumor suppressor of glioblastoma, which not only suppresses tumor growth but also inhibits generation of tumor-initiating cells.

Results

DACH1 Gene on Chromosome 13q21 Is Homozygously Deleted in Glioblastoma.

To identify genomic alterations involved in gliomagenesis, we performed DNA copy number analysis of eight GBMs by using SNP genotyping array (Fig. 1A and Fig. S1A), as well as analysis of a corresponding normal blood DNA for highlighting tumor-specific alterations. We observed high-level amplification at chromosome 7q21 (inferred total copy number >8) and copy number reduction within chromosome arm 4q, 10p, 13q, 16q, 17q, and 18q, and we detected homozygously deleted loci at chromosome 9q21 and 10q23, which spanned known tumor suppressor genes CDKN2A and PTEN, respectively. In addition, we found a unique homozygous deletion at chromosome 13q21 in GBM case 4 and LOH at the region in GBM case 1 and 3 (Fig. 1A and Fig. S1A). Although loss of chromosome 13q14.2 spanning RB1 gene is frequently observed in human malignancies including GBMs (12, 13), the homozygous deletion of chromosome 13q21 has not been reported.

Fig. 1.

DACH1 is homozygously deleted in GBMs. (A) An integrated view of DNA copy numbers and allelic alternations of GBM case 4. Scatter plot of inferred allelic copy numbers (red and blue) was estimated by GIM algorithm. An arrow indicated homozygously deleted loci. (Right) Magnified view of DNA copy number of GBM case 4 from chromosome 13q21 to 13q34. One allele showed DNA copy number reduction in whole region (blue), whereas the other allele (red) showed copy number reduction only at chromosome 13q21. (B) Summary of SNP genotyping in the 13q21 region. S1-24 and M1-11 are SNP ID available in SNP genotyping array and MassARRAY, respectively. Black line, the homozygously deleted region; dotted line, the possible extended region of homozygous deletions; *, also examined by SNP arrays shown in Fig. S1A.

To examine allelic DNA copy numbers at the chromosome 13q21 region in additional GBM cases, we performed targeted genotyping analysis of 28 paired GBMs and blood cells by high-density mass spectrometric analysis using MassARRAY (14, 15). The chromosomal losses were found at least in 11 samples (GBM case 1, 3, 4, 5, 18, 19, 24, 25, 26, 27, and 28; 39.3%), and three of them (GBM case 4, 5, and 27; 10.7% of GBMs) displayed homozygous deletion (Fig. 1B and Fig. S1B). By combination of SNP genotyping array with MassARRAY analysis, homozygously deleted region at chromosome 13q21 of GBM case 4 was restricted to rs1999603 (probe S10)–rs1326684 (probe M8), which might be extended from rs9542598 (probe M2) to rs1421280 (probe S15). In two additional GBM cases, homozygous deletions found by MassARRAY at this locus were from rs10492537 (probe M6) to rs3818437 (probe M7), which might be extended to rs1931443 (probe M3)–rs7332388 (probe M10) in GBM case 5, and from rs10492537 (probe M6) to rs1981186 (probe M9), which might be extended to rs10507796 (probe M4)–rs1886452 (probe M11) in GBM case 27. The boundary of the common homozygously deleted region on the centromeric side was estimated at rs10507796 (probe M4), which was genotyped as LOH in case 27, whereas the telomeric boundary was estimated at rs1421280 (probe S15), which was genotyped as LOH in case 4 (Fig. 1B). Because the homozygously deleted region from rs10507796 to rs1421280 overlapped with DACH1 gene and did not span the adjacent genes to DACH1 such as LOC440145 and KLHL1, we thought DACH1 was a target of these deletions and could be a potential candidate as a tumor suppressor gene of GBMs. Quantitative PCR analysis demonstrated that DNA copy numbers of GBM samples with a homozygous deletion at DACH1-region were much lower than that of whole brain and normal fibroblast cells KMS-6, which has a normal karyotype as (46, XX) (Fig. S2A). In addition, we confirmed protein expression of DACH1 was reduced in glioma cells with a homozygous deletion at DACH1 region, whereas it was detectable in vascular endothelial cells (Fig. S2B).

DACH1 Expression Reduces Growth of Glioma Cells.

We screened an expression level of DACH1 gene in a series of glioma cell lines. Forced expression of lentivirus-carrying DACH1 reduced cell proliferation of SF188, U87MG, T98G, and LNG-308 glioma cell lines, where endogenous expression of DACH1 was not detectable (Fig. S3 A and B). We then generated two U87MG-derived cell lines: U87TR-Da clone-16 and -18, where DACH1 expression could be induced by the addition of doxycycline (Fig. 2A and Fig. S4A). DACH1 decreased viability of the cells at 4 and 6 d after addition of doxycycline (Fig. 2B) and also abrogated anchorage-independent growth of the cells in soft agar (Fig. S3 C and D). We next examined the impact of DACH1 expression on in vivo growth of s.c.-injected tumors of U87TR-Da clone-16 and -18. Increased expression of DACH1 was observed at 8 d after replacement of doxycycline-free drinking water by doxycycline-supplemented water (Fig. 2 C and D). Growth of the U87TR-Da clone-16 and clone-18 tumors was significantly decreased by administration of doxycycline compared with U87TR-Da tumors supplemented with normal drinking water (Fig. 2 E and F and Fig. S4B). Tumor formation was reproduced by stereotactic intracerebral inoculation of DACH1-nonexpressing U87TR-Da cells, whereas tumor was not detectable by the injection of DACH1-expressing U87TR-Da cells (Fig. 2 G and H and Fig. S3E). These results demonstrated that DACH1 decreased growth and proliferation of glioma cells both ex vivo and in vivo, supporting that DACH1 is a tumor suppressor gene of GBMs.

Fig. 2.

DACH1 expression repressed growth of glioma cells. (A) Immunoblotting of U87TR-Da clone-16 and clone-18 cells with an anti-DACH1 antibody. (B) Cell proliferation of U87TR-Da cells, counted by WST-8 assay at 4 or 6 d after induction of DACH1 by doxycycline. (C) Experimental model of tumor progression affected by DACH1 expression. Serially diluted U87TR-Da cells were injected s.c. into the backs of BALB/c nude mice, and the tumor formation was observed at 28 d after s.c. injection of the cells. (D) Expression of DACH1 in U87TR-Da tumor. DACH1 was detected by immunoblotting of xenografted U87TR-Da clone-16 tumors with an anti-DACH1 antibody (Upper). Tumor tissues were resected from mice drinking doxycycline-supplemented (left lane) or control water (right lane). β-actin was detected as a loading control (Lower). (E) Tumor formation of xenografted DACH1-nonexpressing or -expressing U87TR-Da clone-16 cells at 20 d after inoculation. (F) Growth of xenografted U87TR-Da clone-16 or U87TR tumors. DOX, doxycycline (1 μg/mL); points, mean (n = 6); bars, SEM of 6 amimals on per group; N.S., not statistically significant; *P < 0.05 (two-way ANOVA), statistically significant compared with doxycycline minus control. (G) Orthotopic xenograft model for assessing the effect of DACH1 expression on tumor progression. U87TR-Da clone-16 cells were precultured in doxycycline-free (Left) or doxycycline-containing (Right) medium. (H) Size of tumors of the brain stereotactically implanted with DACH1-nonexpressing (DOX-) or -expressing (DOX+) cells at 5 wk after implantation. Bars, SEM of 4 amimals per group; N.S., not statistically significant; *P < 0.05 (two-way ANOVA), statistically significant compared with doxycycline minus control.

DACH1 Inhibits Formation of Tumor-Initiating Spheroids of Glioma Cells.

DACH1 is structurally related to c-Ski and SnoN, which act as transcriptional repressors of the transforming growth factor-β (TGF-β) signaling pathway through the interaction with Smad proteins (16). Previous studies showed that human DACH1 inhibited TGF-β signaling through repressing cyclin D1 (CCND1)expression and decreased proliferation of breast cancer cells (17, 18). We examined whether DACH1 expression affected TGF-β signaling in glioma cells. Induced DACH1 expression reduced the proliferation of the U-373MG cells, which are widely used as a cell model for analyzing TGF-β signaling (19), and repressed TGF-β3–stimulated (CAGA)₉- and p800-luciferase activity (Fig. S5 A–C). However, the expression of cyclin D1, which was proposed as a transcriptional target of DACH1 in breast cancer cell lines (18), was not affected by induced expression of DACH1 (Fig. S5D). Because DACH1 might repress expressions of the other target genes, but except for cyclin D1, we sought DACH1-regulated genes, which could affect growth of glioma cells. By global expression analysis of DACH1-high and -low cells, we found FGF2 expression was repressed by DACH1 (Table S1, Fig. 3A, and Fig. S4A). Because FGF2, also named basic FGF (bFGF), is an essential factor for maintenance of self-renewal of glioma-initiating cells (20, 21), we thought DACH1 affected manintenance of self-renewal of glioma-initiating cells. We discovered DACH1, grown in serum-free neurobasal (NBE) medium (22), blocked spheroid formation of U87TR-Da cells, whereas cell morphology in serum-containing medium was not changed by DACH1 expression (Fig. 3B). Because glioma-initiating cells have been proposed to form spheroids (21), we thought that DACH1-nonexpressing U87TR-Da cells, which formed spheroids in NBE medium, showed high tumorigenicity compared with DACH1-expressing U87TR-Da cells in NBE medium. We confirmed that spheroid of DACH-nonexpressing cells cultured in NBE medium showed high expression of CD133, which has been reported as a marker of cells that are capable of tumor initiation (21), and a neural stem cell marker Nestin (Fig. 3 C and D). We then performed s.c. injection of U87TR-Da cells to examine the relationship between spheroid formation and tumorigenicity of the cells. We found tumor formation of DACH1-nonexpressing U87TR-Da cells (3 of 4 mice with 5 × 10³ cells and all mice with 2 × 10⁴ and 1 × 10⁵ cells), whereas DACH1-expressing U87TR-Da cells did not form any tumors (Fig. 3E). Because DACH1-expressing U87TR-Da cells in serum-containing DMEM did not form tumors with so few cells (2 × 10⁴ to 1 × 10⁵ cells), tumor-initiating cells might be enriched in spheroids of DACH1-nonexpressing U87TR-Da cells grown in serum-free NBE medium.

Fig. 3.

Spheroid formation was inhibited by DACH1 expression. (A) Protein expression levels of DACH1 and bFGF in U87TR-Da clone-16 cells. (B) Images of cultured cells. U87TR-Da clone-16 cells were first cultured in FBS-containing DMEM in the absence (Upper Left) or presence (Lower Left) of doxycycline for 72 h, and then culture medium was replaced into serum-free NBE or serum containing DMEMF medium. Branched projection of adherent cells was observed in DACH1-nonexpressing cells (red arrows). (Scale bars: 500 μm.) (C) Nestin staining of spheroid (green). Nucleus was stained with DAPI (blue). (D) CD133 expression of DACH1-nonexpressing (DOX-) or -expressing (DOX+) cells under serum-containing (DMEMF) or serum-free (NBE) culture condition. Columns, mean (n = 3); bars, S.D.; *, **P < 0.05. (E) Tumor formation of DACH1-expressing or DACH1-nonexpressing U87TR-Da cells cultured in NBE medium. Tumor formation in 3 of 4 mice with 5 × 10³ DACH1-nonexpressing U87TR-Da clone-16 cells and in all mice with 2 × 10⁴ and 1 × 10⁵ DACH1-nonexpressing U87TR-Da clone-16 cells was observed, whereas s.c. injection of DACH1-expressing U87TR-Da clone-16 cells did not form any tumors. DOX-, mice with U87TR-Da clone-16 cells cultured in doxycycline-free NBE medium; DOX+, mice with U87TR-Da cells cultured in doxycycline-supplemented medium.

FGF2 Rescues DACH1-Repressed Tumorigenicity.

FGF2 expression was highly induced under serum-free culture condition, however, DACH1 repressed FGF2 expression at low levels (Fig. 4A). We verified that DACH1 reduced a reporter activity of luciferase cis-regulated by FGF2 promoter (Fig. S6A) and directly bound to FGF2 promoter region in DACH1-expressing U87TR-Da cells cultured in both serum-containing DMEM and serum-free NBE medium by ChIP analysis with an anti-DACH1 antibody (Fig. S6B). Because overexpression of FGF2 was frequently observed in high-grade gliomas and involved in malignant progression of gliomas (23, 24) and a previous study showed that bFGF enhanced tumor-initiating spheroid formation of glioma cells (20), we examined whether spheroid formation of DACH1-expressing U87TR-Da cells was enhanced by exposure of cells to bFGF. DACH1-expressing U87TR-Da cells did not form spheroid, but exogenous bFGF-induced spheroid formation of DACH1-expressing U87TR-Da cells, indicating that bFGF, which was repressed by DACH1, increased the number of spheroid-forming tumor-initiating cells (Fig. 4B and Fig. S4C). Morphology of primary tumor spheroid, which did not express DACH1, was not markedly different under culture conditions with or without bFGF. However, by lentiviral expression of DACH1, the spheroid formation was partially disrupted, and most spheroid-forming cells started to differentiate even under bFGF-supplemented condition (Fig. S4D). We confirmed the reduced expression of a glioma stem cell marker CD133 in primary tumor-derived cells with DACH1 expression under both bFGF2-supplemented and -unsupplemented culture conditions (Fig. S4E). To examine whether bFGF overexpression can confer an increased tumorigenicity, the intracerebral implantation of the cells ectopically overexpressing DACH1 and FGF2 was performed (Fig. 4C). Tumor formation in DACH1-expressing cells with ectopic expression of FGF2 was much higher than that in control DACH1-expressing cells, showing that rescue of FGF2 repression in DACH1-expressing cells increases intracerebral tumor formation (Fig. 4 D and E and Fig. S6C). These results suggested that DACH1 suppress tumor formation through transcriptional repression of FGF2.

Fig. 4.

Ectopic DACH1 expression decreases FGF2 expression, spheroid formation, and tumor growth. (A) Repressed FGF2 expression by DACH1. DMEMF, serum-containing DMEM; NBE, serum-free Neurobasal medium. (B) Spheroid formation in U87TR-Da in the absence and presence of bFGF. Columns, mean (n = 3); bars, SEM of three experiments; *P < 0.05 (unpaired t-test), statistically significant compared with doxycycline minus control. (C) Protein expression of green fluorescent protein (GFP), DACH1, and bFGF. (D) Orthotopic xenograft model for assessing the effect of FGF2 on tumor progression of DACH1-expressing cells. U87MG cells were lentivirally transduced with DACH1 and cultured in zeocin-containing medium. Then, DACH1-expressing cells were infected with lentivirus carrying FGF2 (Left) or control (Right) vector. (E) Size of tumors of the brain implanted with FGF2-expressing or control U87MG cells. Bars, SEM of 5 amimals on per group; N.S., not statistically significant; *P < 0.05 (two-way ANOVA), statistically significant compared with doxycycline minus control.

Discussion

Loss of chromosome 13q has been reported to occur frequently in GBMs (12, 25, 26). Homozygous deletions, LOH, and mutations in RB1 gene, which is located at chromosome 13q14.2, 23 Mb centromeric to DACH1, are found in human cancers including GBMs (13, 26). In this study, we found another locus with a homozygous deletion at 13q21 by SNP genotyping array and targeted genotyping analysis with a mass spectrometer. A previous study reported LOH of chromosome 13q was more common in secondary than in primary glioblastomas (3), and the mutations of IDH1 and IDH2 genes have suggested to associate with development of secondary glioblastomas (27–30). However, because significant relationship between loss of DACH1 and mutations of IDH1 and IDH2 were not found, we think that the mechanisms of DACH1 loss and gliomagenesis categorized into primary or secondary glioblastoma are different (Table S2).

DACH1-spanning region is indicated as a target of copy number variations (CNV), where healthy individuals also exhibit copy number changes. Although it is hard to detect tumor-specific small deletions by a general copy number analysis using an algorithm based on moving window, our GIM algorithm can normalize aberrant copy number change even in CNV region by using signal data of both tumor DNA and corresponding normal DNA (4, 11). We overcame CNV effect and could detect small homozygous deletions at DACH1 region by an improved algorithm for SNP genotyping array with the targeted genotyping analysis using a mass spectrometer.

To explore additional mechanisms for DACH1 inactivation other than genomic loss, we first examined sequencing of DACH1 exonic regions by using genomic DNA from 25 GBM tumors and 8 glioma cell lines and did not observe any somatic mutations, whereas GBM case 25 showed LOH and the remained allele with nonsynonymous polymorphism, which could be translated into amino acid-substituted DACH1, V464M (Fig. S2C). It should require further examination such as loss-of-function of the protein and frequency of mutational events with amino acid substitution of DACH1 protein in GBMs. We then examined methylation status of DACH1 promoter region, because a CpG island existed in the vicinity of transcription start site (TSS) of DACH1. Because DNA hypermethylation at DACH1 promoter region was frequently observed (Fig. S2D), hypermethylation of the DACH1 promoter could be an additional mechanism for DACH1 inactivation in gliomagenesis.

Homozygous deletions observed in all three cases contained a deduced promoter region and the first exon with ATG start codon of DACH1 gene, indicating the homozygous deletions cause functional disruption of DACH1. Forced expression of DACH1 decreases growth of glioma cells and inhibits formation of spheroids, which are proposed as a tumor-initiating cell population and chemoresistant population of glioma cells. Even though downstream signaling of bFGF is well described, only a few reports have provided the mechanism of transcriptional regulation of FGF2 (31, 32). Here, we demonstrated that DACH1, a member of corepressor complexes, repressed FGF2 expression. Campanelli et al. found the expression level of DACH1 in glial precursors was much higher than that in neural stem cells (NSCs) (33). Withdrawal of bFGF, which is known as a critical component of culture medium for human embryonic and neural stem cells (34), drives astrocytic differentiation (2, 34, 35). So we think that loss or mutation of DACH1 may interrupt astrocytic differentiation of glial-restricted progenitor cells through sustained expression of bFGF and, subsequently, trigger gliomagenesis by convertion of progenitor cells to tumor-initiating cells.

Although loss of DACH1 was detected in some GBM specimens and glioma cell lines at genomic and/or transcriptional level, there were other glioma cell lines that exhibited detectable DACH1 expression, such as U251MG (Fig. S3A). When we silenced DACH1 expression of U251MG by RNA interference (RNAi) (Fig. S7A), slightly increased cell proliferation was observed (Fig. S7B). FGF2 and CD133 expression was slightly up-regulated when DACH1 expression was silenced by RNAi (Fig. S7 C and D). We also observed increased spheroid formation in DACH1-silenced cells under serum-free culture condition (Fig. S7E). However, these effects of DACH1 silencing in U251MG cells were relatively slight. Because U251MG exhibited cell-autonomous growth even with high DACH1 expression, there must be additional alterations in oncogenes/tumor suppressor genes other than DACH1, which might obscure the effect of DACH1 silencing on tumor suppressing activity.

In the analysis with expression microarrays, we observed that DACH1 decreased expression of secreted factors including TGF-β2, leukemia inhibitory factor (LIF), and interleukin-6 (IL-6) (Table S1). Recent reports demonstrated that TGF-β enhanced the self-renewal property of glioma-initiating cells by activating LIF-JAK/STAT pathway (36), and autocrine TGF-β signaling is involved in maintenance of tumorigenicity of glioma-initiating cells (19). Both LIF and IL-6 also are activators of JAK/STAT pathway. So we presumed that DACH1 may suppress tumorigenesis through repression of not only FGF2 but also the above tumor-initiating factors. Potential tumor-initiating factors, such as bFGF, TGF-β, LIF, and IL-6, will also be good candidates drug targets for GBMs.

The previous integrated genomic analysis of glioblastoma identified genomic alterations in genes belonging not only to a variety of cellular process pathways, which were likewise altered in many types of cancers, but also to nervous system-specific cellular pathways (27). From our results, we think that DACH1 may function as one of tumor suppressor genes regulating through nervous system-specific cellular pathways. Little is yet known about the origin of glioblastoma cells, but some reports suggested that NSCs in the subventricular zone (SVZ) of the adult brain might be a candidate origin of cancer stem cells of glioblastoma (37, 38). Although we could not detect DACH1 expression at adult SVZ, a high expression of DACH1 was observed at the ventricular wall of fetus (Fig. S2 E and F), implicating involvement of DACH1 in maturation of neural cells. Li and colleagues demonstrated murine Dach1 regulated retinogenesis and pituitary development through tissue-specific gene regulation by recruiting the corepressors (39, 40). Functional relevance of DACH1 expression in development and neural differentiation of both invertebrates and vertebrates has been reported (39–46). DACH1-expressed embryonic and postnatal brain-derived cells displayed neural stem cell-like property (44), suggesting that DACH1 functions in neural differentiation. From these facts together, we think that DACH1 act as a guardian of differentiation in the glial lineage, and loss of DACH1 would result in dysregulation of normal differentiation and drive gliomagenesis.

Taking advantages of our allelic DNA copy number analysis for tumors and corresponding normal cells by using SNP genotyping array as well as MassARRAY, we provide evidence that DACH1 is homozygously deleted in GBMs. As far as we know, this is the first study showing loss of DACH1 gene function in tumor cells at the genomic level. Our observation supports recent studies indicating that DACH1 scarcely expressed in tumor-initiating cells and such low expression correlated with poor prognosis of breast cancers (17, 18, 47). We showed that DACH1 expression decreases proliferation of glioma cells and suppresses tumorigenicity through inhibition of bFGF-dependent spheroid formation. We here propose DACH1 is a unique tumor suppressor gene, which does not only suppresses tumor growth but also inhibits generation of tumor-initiating cells during neural differentiation. Understanding molecular basis of DACH1-mediated epigenetic regulation may provide mechanism of both neural differentiation and gliomagenesis.

Methods

Materials.

All clinical samples were obtained with the informed consent of the patients after permission by the ethics committees of Tokyo University Hospital. Tumors were diagnosed according to World Health Organization classification (48).

SNP Genotyping Array.

SNPs of peripheral blood cells or GBM samples were genotyped by 50K Xba SNP mapping arrays (Affymetrix) according to GeneChip Mapping 50K Assay Manual (4). Allelic and total DNA copy numbers were calculated by GIM algorithm (11).

MassARRAY Analysis.

SNP genotyping of 28 GBM samples, including the 8 samples used in the initial screening by SNP mapping array analysis, was performed with MassARRAY Genotyping system (Sequenom). Paired DNA samples from blood or tumor were genotyped in duplicate. Quantitation of the peak area was performed by Sequenom’s MassARRAY RT software. The threshold for LOH was defined as 40% reduction of one allele in tumor sample, as described (14).

Gene Expression Analysis.

Total RNA was extracted with TRIzol (Life Technologies). Biotin-labeled cRNA was synthesized and hybridized to GeneChip U133 plus 2 (Affymetrix) according to the manufacturer's instruction. Gene expression data were normalized by using the MAS5 algorithm. Data were deposited in National Center for Biotechnology Information Gene Expression Omnibus (GEO) database (GEO accession no. GSE19678; www.ncbi.nlm.nih.gov/geo/).

Immunoblotting.

An anti-DACH1 antibody was generated by immunization of GST-fused DACH1 protein and purified by affinity-purification with Immunogen-bound beads. Immunoblot was performed by using anti-DACH1 or anti–β-actin (AC-40; Sigma) antibody according to a previous report (49). Cell lysates at 3 d after the addition of doxycycline or RNAi treatment were used.

Spheroid Formation Assay.

Expression of DACH1 was induced by exposure of U87TR-Da-clone16 or -clone18 cells to doxycycline for 72 h and then cells were grown in NBE medium (22, 50) composed of Neurobasal Medium (Life Technologies), N-2, B-27 supplement (0.5× each; Life Technologies), human recombinant bFGF (50 ng/mL; PeproTech), and EGF (50 ng/mL; Life Technologies). At 10 d after medium change, the number of spheroid-forming colonies was counted. Paraformaldehyde-fixed spheroids were stained with an anti-Nestin monoclonal antibody (BD Biosciences) and Alexa488-conjugated anti-mouse IgG antibody (Life Technologies).

In Vivo Studies.

Male 6- to 8-wk-old male athymic (nude) mice (nu/nu) of BALB/c were purchased from Charles River Laboratory Japan. For assessment of in vivo tumor growth, U87TR-Da cells were injected s.c. into the flank of mice. At the start point (day 0), drinking water was replaced with doxycycline-containing water for induction of DACH1 expression. For assessment of tumorigenicity, before injection, DACH1 expression was induced by the exposure of the U87TR-Da clone-16 cells with doxycycline for 10 d. The tumor formation was observed after s.c. injection of the serially diluted cells. For intracerebral stereotactic inoculation, cells in 6 μL of PBS were implanted into the right corpus striatum of the anesthetized mouse brain. Animal studies were performed according to institutional guidelines. See details of in vivo tumor growth and tumorigenicity assay in SI Methods.

Full methods and any associated references are available in SI Methods. Primers used were listed in Table S3.